Patterns of multi-morbidity and prediction of hospitalisation and all-cause mortality in advanced age

Patterns of multi-morbidity and prediction of hospitalisation and all-cause mortality in advanced... Abstract Background multi-morbidity is associated with poor outcomes and increased healthcare utilisation. We aim to identify multi-morbidity patterns and associations with potentially inappropriate prescribing (PIP), subsequent hospitalisation and mortality in octogenarians. Methods life and Living in Advanced Age; a Cohort Study in New Zealand (LiLACS NZ) examined health outcomes of 421 Māori (indigenous to New Zealand), aged 80–90 and 516 non-Māori, aged 85 years in 2010. Presence of 14 chronic conditions was ascertained from self-report, general practice and hospitalisation records and physical assessments. Agglomerative hierarchical cluster analysis identified clusters of participants with co-existing conditions. Multivariate regression models examined the associations between clusters and PIP, 48-month hospitalisations and mortality. Results six clusters were identified for Māori and non-Māori, respectively. The associations between clusters and outcomes differed between Māori and non-Māori. In Māori, those in the complex multi-morbidity cluster had the highest prevalence of inappropriately prescribed medications and in cluster ‘diabetes’ (20% of sample) had higher risk of hospitalisation and mortality at 48-month follow-up. In non-Māori, those in the ‘depression-arthritis’ (17% of the sample) cluster had both highest prevalence of inappropriate medications and risk of hospitalisation and mortality. Conclusions in octogenarians, hospitalisation and mortality are better predicted by profiles of clusters of conditions rather than the presence or absence of a specific condition. Further research is required to determine if the cluster approach can be used to target patients to optimise resource allocation and improve outcomes. comorbidity, hospitalisation, aged, indigenous population, older people Introduction Multi-morbidity increases with age [1, 2] being present in up to 82% of individuals aged over 85 years [1]. Multi-morbidity is associated with: rapid functional decline amongst older people [3]; memory problems [4]; lower quality of life (QoL) [5], and is a stronger predictor of healthcare utilisation than age or gender [6]. The ageing of the population, with those aged 85+ being both the fastest growing group and the most likely to have multi-morbidity, will mean understanding and managing multi-morbidity in the oldest old is increasingly important. Studies of multi-morbidity lack consistency of methods [7]. Patterns of diseases may be more useful than counting number of conditions. While patterns of multi-morbidity have been identified [8], studies to date seldom link outcomes to patterns, and none have examined octogenarians or indigenous older populations. New understandings of outcomes related to specific disease groupings may lead to new strategies in improving patient-centred care and outcomes. We aim to identify specific patterns of conditions of octogenarians living in NZ, contrasting two ethnic groups, and report associated prescribing appropriateness and subsequent hospitalisations and mortality over 4 years. Methods Te Puāwaitanga O Nga Tapuwae Kia ora Tonu: Life and Living in Advanced Age; a Cohort Study in NZ (LiLACS NZ) is a cohort study of Māori (indigenous people in New Zealand) and non-Māori octogenarians [9]. The study recruited a population based sample of 421 Māori aged 80–90 years and 516 non-Māori aged 85 years between March 2010 and April 2011 from a regionally defined area in the North Island, NZ. Detailed recruitment strategies are previously described [10]. The study was approved by the Northern X Regional Ethics Committee (NXT/09/09/088; NXT/10/12/127). Written informed consent was obtained from all participants. Data collection Socio-economic, health, and physical status were established using standardised techniques (comprehensive face-to-face interviews and physical assessments). For participants who scored <73 in the Modified Mini Mental Status Examination, a proxy was identified and completed the interview (details in Supplementary material). Diagnoses Fourteen pre-specified medical conditions prevalent in older adults were identified from self-report (‘Have you ever been told by a doctor that you have had [condition]?’), General Practice (GP) records (list of conditions), hospitalisation records, physical assessments (Figure 1). Only participants having a record of presence/absence of respective conditions in one of the five sources were included in analyses [11]. Figure 1. View largeDownload slide Data sources for participants in LiLACS NZ. LILACS NZ, Life and Living in Advanced Age; a Cohort Study in New Zealand; GP, general practitioner; NZHIS, New Zealand Health Identification System aQuestionnaire: 5 participants did not complete questionnaire. bGP medical records: No consent, n = 41. Consented but review was not completed (n = 59), Consented but participants changed their mind (n = 24), Consented but GP refused to give out information (n = 2). cNZHIS: 60 participants refused consent. dPhysical assessments: No consent, n = 306. Of those who consented (n = 630), 624 had blood pressure measurements, six did not have BP measurements (participants refused, n = 3; unable to do, n = 2, participant out of town); 615 had an ECG read by cardiologist, 15 ECGs were not received (participants refused ECG n = 4; unable to physically co-operate, n = 3; unsuitable environment n = 2; other reasons n = 6). eBlood test: No consent, n = 305. Consented but did not gave sample, n = 76. Of those who consented (n = 556), 546 had glucose analysis, 10 did not have glucose analysis due to laboratory oversight. Figure 1. View largeDownload slide Data sources for participants in LiLACS NZ. LILACS NZ, Life and Living in Advanced Age; a Cohort Study in New Zealand; GP, general practitioner; NZHIS, New Zealand Health Identification System aQuestionnaire: 5 participants did not complete questionnaire. bGP medical records: No consent, n = 41. Consented but review was not completed (n = 59), Consented but participants changed their mind (n = 24), Consented but GP refused to give out information (n = 2). cNZHIS: 60 participants refused consent. dPhysical assessments: No consent, n = 306. Of those who consented (n = 630), 624 had blood pressure measurements, six did not have BP measurements (participants refused, n = 3; unable to do, n = 2, participant out of town); 615 had an ECG read by cardiologist, 15 ECGs were not received (participants refused ECG n = 4; unable to physically co-operate, n = 3; unsuitable environment n = 2; other reasons n = 6). eBlood test: No consent, n = 305. Consented but did not gave sample, n = 76. Of those who consented (n = 556), 546 had glucose analysis, 10 did not have glucose analysis due to laboratory oversight. Outcomes Medication use was established by direct observation of medications (prescribed and non-prescribed), participants were asked how it was taken and were coded with WHO-ATC codes [12]. Prescribing quality was assessed using STOPP/START criteria [13]. STOPP identifies potentially inappropriate medications (PIMs); START lists potential prescribing omissions (PPOs). Hospitalisations and mortality were ascertained from NZ Ministry of Health data based on individually matched National Health Identification number for 48 months post-enrolment. Statistical analysis We chose to examine Māori and non-Māori separately because we know that prevalence of diseases [14] and outcomes from these conditions differ between Māori and non-Māori [15, 16], and different ethnic groups may need different interventions strategies [17]. Patient-centred approaches necessitate patient-centred rather than condition-centred analyses. Agglomerative hierarchical cluster analysis was used to identify clinically relevant subgroups based on groupings of co-existing conditions on all participants with complete data. An example of this method is species taxonomy. For the clustering algorithm Ward’s method was used. This sorts individuals into clusters containing a similar profile of conditions [18]. The number of clusters was guided by a combination of methods including visualising the dendrograms and scree plots. Determining the best number of clusters is a balance between manageability and strategic attention. We adapted Cornell’s pre-defined criteria [18], meaning the cluster solution was based on informed judgements and consensus of health researchers (geriatricians, cardiologist, general practitioner, biostatisticians, epidemiologist). Stability of the cluster solution was assessed by repeating the process on a randomly selecting 50% of the Māori and non-Māori sample. Cluster labelling was based on the most prevalent conditions within the cluster and highest representation of the condition across clusters. Descriptive statistics are presented for each of the clusters. Multivariate regression models (linear or log-binomial regression) examined associations between identified clusters and PIMs, PPOs, and pre 12-month hospitalisation adjusting for age (Māori only), gender, education status and deprivation index. Cox-proportional hazards models were used to determine hazard ratio (HR) for 48-month hospitalisation and all-cause mortality. We used the cluster identified as ‘healthiest’ (fewer conditions or lower prevalence of all conditions versus other clusters) as the reference group in regression analyses. Separate analyses (adjusted for similar confounders) were completed based on median number of conditions. Statistical analyses were performed with IBM SPSS v23. A P-value < 0.05, two-tailed, was considered statistically significant. Results Participant characteristics In the study, 888/937 participants were included in analyses (387/421 Māori; 501/516 non-Māori) (Figure 1). The proportion of participants living in residential care was similar between those included and excluded (8% vs. 13%, P = 0.176), as with functional status (NEADL score), median(IQR) 19(4) vs. 17(5), P = 0.589. Among the 387 Māori, 221 (57%) were women; mean age was 82.7 years (SD 2.8). All the 501 non-Māori were 85 years old at entry; 268 (54%) were women. The median(IQR) for number of conditions overall was 3(3) for both Māori and non-Māori. Inappropriate medication was identified in 168 (66%) Māori and 250 (62%) non-Māori. Over 48-month follow-up, all-cause hospitalisations occurred in 302 (80%) Māori and 391 (79%) non-Māori; 121 (31%) Māori and 133 (27%) non-Māori died. For Māori, those with ≥3 conditions had higher risk of any hospitalisation (P < 0.01) and all-cause mortality (P < 0.001) at 48-month follow-up than those with <3 conditions (Table 1). Non-Māori with ≥3 conditions had a higher risk of all-cause mortality (P < 0.001) at 48-month follow-up (Table 2). The risk of any hospitalisation did not differ between those with less than three conditions and those having three or more. Table 1. Prevalence, patterns of conditions and outcomes for Māori   Cluster 1, 73/387 (19%)  Cluster 2, 59/387 (15%)  Cluster 3, 45/387 (12%)  Cluster 4, 98/387 (25%)  Cluster 5, 79/387 (20%)  Cluster 6, 33/387 (9%)  ‘Well’ (lower prevalence of conditions)  CHF and AF  Arthritis  CVD, respiratory and mental health  Diabetes  Complex multi-morbidity (high prevalence of most conditions)    Mean (SD)  Age  82.5 (2.9)  83.2 (3.0)  83.3 (2.5)  82.8 (2.7)  82.4 (2.5)  82.4 (2.8)  Number of conditions  1.1 (1.3)  3.7 (1.4)  2.7 (1.2)  3.4 (1.9)  3.8 (1.7)  6.5 (1.4)    n (%)  Sex, male  30 (41)  31 (53)  13 (29)  41 (42)  37 (47)  14 (42)  Deprivation Index     Low  8 (11)  11 (19)  9 (20)  13 (13)  11 (14)  5 (15)   Medium  25 (34)  18 (30)  9 (20)  23 (24)  21 (27)  6 (18)   High  40 (55)  30 (51)  27 (60)  62 (63)  47 (59)  22 (67)  CAD/PVD  1 (1)  41 (69)  14 (31)  71 (72)  47 (59)  31 (94)  CHF  1 (1)  43 (73)  4 (9)  15 (15)  39 (49)  22 (67)  Stroke/TIA  4 (5)  13 (22)  8 (18)  43 (44)  20 (25)  7 (21)  Any AF  3 (4)  52 (88)  5 (11)  7 (7)  30 (38)  21 (64)  DM  2 (3)  0 (0)  9 (20)  17 (17)  75 (95)  15 (45)  Asthma or COPD  7 (10)  14 (24)  1 (2)  62 (63)  21 (27)  30 (91)  Osteoporosis  0 (0)  4 (7)  21 (47)  15 (15)  6 (8)  16 (48)  Osteoarthritis  5 (7)  9 (15)  38 (84)  23 (23)  4 (5)  31 (94)  RA  13 (18)  2 (3)  7 (16)  10 (10)  13 (16)  24 (73)  Dementia  12 (16)  11 (19)  6 (13)  23 (23)  11 (14)  0 (0)  Depression  12 (16)  17 (29)  6 (13)  37 (38)  15 (19)  9 (27)  Thyroid disease  3 (4)  0 (0)  0 (0)  4 (4)  4 (5)  3 (9)  Non-skin cancer  13 (18)  11 (19)  1 (2)  8 (8)  17 (22)  6 (18)  Melanoma  2 (3)  0 (0)  0 (0)  1 (1)  0 (0)  0 (0)    Cross-sectional associations with cluster membership Mean (SE); Beta (95%CI) for adjusted model  Prescribed medications  2.4 (0.4); Reference  5.6 (0.5); 3.25 (2.03–4.47)**  3.6 (0.5); 1.25 (0.04–2.46)*  5.1 (0.4); 2.68 (1.65–3.71)**  6.2 (0.4); 3.84 (2.76–4.92)**  7.9 (0.5); 5.54 (4.30–6.79)**    n (%); OR (95%CI) for adjusted model  PIMS  6 (11%); Reference  15 (45%); 4.05 (1.72–9.56)**  5 (15%); 1.45 (0.47–4.45)  8 (14%); 1.20 (0.43–3.31)  12 (25%); 2.05 (0.83–5.06)  15 (52%); 4.51 (1.94–10.50)**  PPOS  11 (21%); Reference  16 (49%); 2.57 (1.34–4.96)**  22 (65%); 3.56 (1.97–6.45)**  40 (69%); 3.52 (1.99–6.22)**  36 (74%); 3.87 (2.19–6.83)**  25 (86%); 4.32 (2.45–7.61)**  Pre 12-month admission  10 (14%); Reference  26 (45%); 3.23 (1.63–6.41)**  11 (25%); 2.20 (1.00–5.00)  37 (39%); 3.00 (1.56–5.81)**  35 (45%); 3.46 (1.81–6.65)**  21 (66%); 4.82 (2.48–9.38)**    Longitudinal outcomes of cluster membership n (%); HR (95%CI) for adjusted model  48-month any hospitalisationa  48 (67%); Reference  48 (83%); 1.41 (0.93–2.15)  36 (82%): 1.77 (1.12–2.79)*  74 (78%): 1.29 (0.88–1.89)  67 (86%); 1.51 (1.02–2.22)*  29 (91%); 1.70 (1.05–2.77)*  48-month mortalityb  13 (18%): Reference  16 (27%); 1.47 (0.68–3.15)  14 (31%); 1.56 (0.70–3.50)  33 (34%); 2.02 (1.03–3.95)*  32 (41%); 2.58 (1.32–5.05)**  13 (39%); 2.58 (1.17–5.66)*    Longitudinal outcomes of number of conditions    <3 conditions, 147/387  ≥3 conditions, 240/387  48-month any hospitalisationa  100 (70%); Reference  202 (85%); 1.54 (1.19–1.99)**  48-month mortalityb  27 (18%); Reference  94 (39%): 2.46 (1.58–3.83)**    Cluster 1, 73/387 (19%)  Cluster 2, 59/387 (15%)  Cluster 3, 45/387 (12%)  Cluster 4, 98/387 (25%)  Cluster 5, 79/387 (20%)  Cluster 6, 33/387 (9%)  ‘Well’ (lower prevalence of conditions)  CHF and AF  Arthritis  CVD, respiratory and mental health  Diabetes  Complex multi-morbidity (high prevalence of most conditions)    Mean (SD)  Age  82.5 (2.9)  83.2 (3.0)  83.3 (2.5)  82.8 (2.7)  82.4 (2.5)  82.4 (2.8)  Number of conditions  1.1 (1.3)  3.7 (1.4)  2.7 (1.2)  3.4 (1.9)  3.8 (1.7)  6.5 (1.4)    n (%)  Sex, male  30 (41)  31 (53)  13 (29)  41 (42)  37 (47)  14 (42)  Deprivation Index     Low  8 (11)  11 (19)  9 (20)  13 (13)  11 (14)  5 (15)   Medium  25 (34)  18 (30)  9 (20)  23 (24)  21 (27)  6 (18)   High  40 (55)  30 (51)  27 (60)  62 (63)  47 (59)  22 (67)  CAD/PVD  1 (1)  41 (69)  14 (31)  71 (72)  47 (59)  31 (94)  CHF  1 (1)  43 (73)  4 (9)  15 (15)  39 (49)  22 (67)  Stroke/TIA  4 (5)  13 (22)  8 (18)  43 (44)  20 (25)  7 (21)  Any AF  3 (4)  52 (88)  5 (11)  7 (7)  30 (38)  21 (64)  DM  2 (3)  0 (0)  9 (20)  17 (17)  75 (95)  15 (45)  Asthma or COPD  7 (10)  14 (24)  1 (2)  62 (63)  21 (27)  30 (91)  Osteoporosis  0 (0)  4 (7)  21 (47)  15 (15)  6 (8)  16 (48)  Osteoarthritis  5 (7)  9 (15)  38 (84)  23 (23)  4 (5)  31 (94)  RA  13 (18)  2 (3)  7 (16)  10 (10)  13 (16)  24 (73)  Dementia  12 (16)  11 (19)  6 (13)  23 (23)  11 (14)  0 (0)  Depression  12 (16)  17 (29)  6 (13)  37 (38)  15 (19)  9 (27)  Thyroid disease  3 (4)  0 (0)  0 (0)  4 (4)  4 (5)  3 (9)  Non-skin cancer  13 (18)  11 (19)  1 (2)  8 (8)  17 (22)  6 (18)  Melanoma  2 (3)  0 (0)  0 (0)  1 (1)  0 (0)  0 (0)    Cross-sectional associations with cluster membership Mean (SE); Beta (95%CI) for adjusted model  Prescribed medications  2.4 (0.4); Reference  5.6 (0.5); 3.25 (2.03–4.47)**  3.6 (0.5); 1.25 (0.04–2.46)*  5.1 (0.4); 2.68 (1.65–3.71)**  6.2 (0.4); 3.84 (2.76–4.92)**  7.9 (0.5); 5.54 (4.30–6.79)**    n (%); OR (95%CI) for adjusted model  PIMS  6 (11%); Reference  15 (45%); 4.05 (1.72–9.56)**  5 (15%); 1.45 (0.47–4.45)  8 (14%); 1.20 (0.43–3.31)  12 (25%); 2.05 (0.83–5.06)  15 (52%); 4.51 (1.94–10.50)**  PPOS  11 (21%); Reference  16 (49%); 2.57 (1.34–4.96)**  22 (65%); 3.56 (1.97–6.45)**  40 (69%); 3.52 (1.99–6.22)**  36 (74%); 3.87 (2.19–6.83)**  25 (86%); 4.32 (2.45–7.61)**  Pre 12-month admission  10 (14%); Reference  26 (45%); 3.23 (1.63–6.41)**  11 (25%); 2.20 (1.00–5.00)  37 (39%); 3.00 (1.56–5.81)**  35 (45%); 3.46 (1.81–6.65)**  21 (66%); 4.82 (2.48–9.38)**    Longitudinal outcomes of cluster membership n (%); HR (95%CI) for adjusted model  48-month any hospitalisationa  48 (67%); Reference  48 (83%); 1.41 (0.93–2.15)  36 (82%): 1.77 (1.12–2.79)*  74 (78%): 1.29 (0.88–1.89)  67 (86%); 1.51 (1.02–2.22)*  29 (91%); 1.70 (1.05–2.77)*  48-month mortalityb  13 (18%): Reference  16 (27%); 1.47 (0.68–3.15)  14 (31%); 1.56 (0.70–3.50)  33 (34%); 2.02 (1.03–3.95)*  32 (41%); 2.58 (1.32–5.05)**  13 (39%); 2.58 (1.17–5.66)*    Longitudinal outcomes of number of conditions    <3 conditions, 147/387  ≥3 conditions, 240/387  48-month any hospitalisationa  100 (70%); Reference  202 (85%); 1.54 (1.19–1.99)**  48-month mortalityb  27 (18%); Reference  94 (39%): 2.46 (1.58–3.83)**  CAD, coronary artery disease; CHF, congestive heart failure; TIA, transient ischaemic attack; PVD, peripheral vascular disease; DM, Type II diabetes mellitus; AF, atrial fibrillation; COPD, chronic obstructive pulmonary disease; RA, rheumatoid arthritis; CVD, cardiovascular disease; SE, standard error; CI, confidence interval. Missing data: Number for prescribed medications, PIMs, PPOs: 130/387 (34%); hospital admission: 8/387 (2%). The bold values highlight prevalences of conditions, not stats significance values. Note: The prevalence of hypertension in Clusters 1 to 6: 87%, 97%, 82%, 87%, 75%, 94%. Base model to examine longitudinal outcomes includes age, gender, education, NZ deprivation index. aBase model plus pre-hospital admission, weighted survival time; bBase model plus 48-month admission. *P < 0.05; **P < 0.01. Table 2. Prevalence and patterns of conditions and outcomes for non-Māori   Cluster 1, 89/501 (18%)  Cluster 2, 66/501 (13%)  Cluster 3, 83/501 (17%)  Cluster 4, 63/501 (13%)  Cluster 5, 111/501 (22%)  Cluster 6, 89/501 (18%)  ‘Well’ (lower prevalence of conditions)  CHF and AF  Depression and arthritis  Cancer  Respiratory and diabetes  Stroke    Mean (SD)  Age  All non-Māori participants born in the same year, 84.6 (0.6)  Number of conditions  0.9 (1.1)  3.2 (1.4)  3.3 (1.6)  2.9 (1.3)  3.0 (1.7)  4.0 (2.1)    n (%)  Sex, male  34 (38)  41 (62)  37 (45)  36 (57)  51 (46)  34 (38)  Deprivation index     Low  31 (35)  18 (27)  11 (13)  17 (27)  32 (21)  23 (26)   Medium  37 (42)  27 (41)  33 (40)  22 (35)  51 (46)  34 (38)   High  21 (24)  21 (32)  39 (47)  24 (38)  37 (33)  32 (36)  CAD/PVD  3 (3)  48 (73)  27 (33)  30 (48)  69 (62)  58 (65)  CHF  3 (3)  30 (45)  7 (8)  10 (16)  30 (27)  27 (30)  Stroke/TIA  1 (1)  5 (8)  17 (20)  3 (5)  26 (23)  84 (94)  Any AF  3 (3)  50 (76)  5 (6)  5 (8)  15 (14)  26 (29)  DM  0 (0)  13 (20)  16 (19)  14 (22)  30 (27)  6 (7)  Asthma or COPD  0 (0)  7 (11)  19 (23)  12 (19)  80 (72)  16 (18)  Osteoporosis  27 (30)  2 (3)  24 (29)  9 (14)  18 (16)  25 (28)  Osteoarthritis  27 (30)  30 (45)  47 (57)  15 (24)  34 (31)  46 (52)  RA  0 (0)  3 (5)  34 (41)  0 (0)  12 (11)  2 (2)  Dementia  10 (11)  8 (12)  16 (19)  1 (2)  7 (6)  5 (6)  Depression  4 (4)  10 (15)  61 (73)  10 (16)  1 (1)  37 (42)  Thyroid disease  4 (4)  3 (5)  4 (5)  6 (10)  7 (6)  13 (15)  Non-skin cancer  0 (0)  5 (8)  6 (7)  59 (94)  4 (4)  16 (18)  Melanoma  4 (4)  3 (5)  5 (6)  11 (17)  6 (5)  2 (2)    Cross-sectional associations with cluster membership Mean (SE); Beta (95%CI) for adjusted model  Prescribed medications, n  3.1 (0.4); Reference  5.5 (0.4); 2.50 (1.37–3.63)**  5.6 (0.4); 2.58 (1.48–3.68)**  4.4 (0.4); 1.31 (0.17–2.45)*  5.6 (0.3); 2.53 (1.55–3.51)**  6.5 (0.4); 3.48 (2.44–4.53)**    n (%); OR (95%CI) for adjusted model  PIMS  15 (20%); Reference  13 (25%); 1.26 (0.66–2.43)  24 (39%); 1.80 (1.04–3.10)*  11 (21%); 1.03 (0.52–2.03)  29 (32%); 1.55 (0.90–2.65)  21 (31%); 1.56 (0.88–2.76)  PPOS  23 (30%); Reference  25 (48%); 1.58 (1.02–2.45)*  38 (62%); 2.00 (1.34–2.97)**  22 (42%); 1.38 (0.86–2.21)  49 (53%); 1.76 (1.19–2.59)**  40 (59%); 1.85 (1.24–2.74)**  Pre 12-month admission  15 (17%); Reference  24 (36%); 1.98 (1.13–3.47)*  25 (30%); 1.68 (0.95–2.96)  18 (29%); 1.59 (0.87–2.91)*  36 (33%); 1.82 (1.06–3.11)*  44 (49%); 2.72 (1.63–4.53)**    Longitudinal outcomes of cluster membership n (%); HR (95%CI) for adjusted model  48-month any hospitalisationa  57 (65%); Reference  55 (83%); 1.32 (0.90–1.94)  71 (86%) 1.48 (1.03–2.12)*  48 (76%); 1.11 (0.75–1.65)  90 (82%); 1.30 (0.92–1.83)  70 (79%); 1.34 (0.94–1.92)  48-month mortalityb  11 (12%); Reference  17 (26%); 2.15 (0.98–4.74)  26 (31%); 2.66 (1.26–5.59)*  18 (29%); 2.32 (1.06–5.09)*  33 (30%); 2.55 (1.24–5.25)*  28 (32%); 2.77 (1.34–5.73)**    Longitudinal outcomes of number of conditions    <3 conditions, 231/501  ≥3 conditions, 270/501  48-month any hospitalisationa  168 (73%): Reference  223 (83%); 1.10 (0.88–1.36)  48-month mortalityb  38 (16%): Reference  95 (35%); 2.24 (1.52–3.29)**    Cluster 1, 89/501 (18%)  Cluster 2, 66/501 (13%)  Cluster 3, 83/501 (17%)  Cluster 4, 63/501 (13%)  Cluster 5, 111/501 (22%)  Cluster 6, 89/501 (18%)  ‘Well’ (lower prevalence of conditions)  CHF and AF  Depression and arthritis  Cancer  Respiratory and diabetes  Stroke    Mean (SD)  Age  All non-Māori participants born in the same year, 84.6 (0.6)  Number of conditions  0.9 (1.1)  3.2 (1.4)  3.3 (1.6)  2.9 (1.3)  3.0 (1.7)  4.0 (2.1)    n (%)  Sex, male  34 (38)  41 (62)  37 (45)  36 (57)  51 (46)  34 (38)  Deprivation index     Low  31 (35)  18 (27)  11 (13)  17 (27)  32 (21)  23 (26)   Medium  37 (42)  27 (41)  33 (40)  22 (35)  51 (46)  34 (38)   High  21 (24)  21 (32)  39 (47)  24 (38)  37 (33)  32 (36)  CAD/PVD  3 (3)  48 (73)  27 (33)  30 (48)  69 (62)  58 (65)  CHF  3 (3)  30 (45)  7 (8)  10 (16)  30 (27)  27 (30)  Stroke/TIA  1 (1)  5 (8)  17 (20)  3 (5)  26 (23)  84 (94)  Any AF  3 (3)  50 (76)  5 (6)  5 (8)  15 (14)  26 (29)  DM  0 (0)  13 (20)  16 (19)  14 (22)  30 (27)  6 (7)  Asthma or COPD  0 (0)  7 (11)  19 (23)  12 (19)  80 (72)  16 (18)  Osteoporosis  27 (30)  2 (3)  24 (29)  9 (14)  18 (16)  25 (28)  Osteoarthritis  27 (30)  30 (45)  47 (57)  15 (24)  34 (31)  46 (52)  RA  0 (0)  3 (5)  34 (41)  0 (0)  12 (11)  2 (2)  Dementia  10 (11)  8 (12)  16 (19)  1 (2)  7 (6)  5 (6)  Depression  4 (4)  10 (15)  61 (73)  10 (16)  1 (1)  37 (42)  Thyroid disease  4 (4)  3 (5)  4 (5)  6 (10)  7 (6)  13 (15)  Non-skin cancer  0 (0)  5 (8)  6 (7)  59 (94)  4 (4)  16 (18)  Melanoma  4 (4)  3 (5)  5 (6)  11 (17)  6 (5)  2 (2)    Cross-sectional associations with cluster membership Mean (SE); Beta (95%CI) for adjusted model  Prescribed medications, n  3.1 (0.4); Reference  5.5 (0.4); 2.50 (1.37–3.63)**  5.6 (0.4); 2.58 (1.48–3.68)**  4.4 (0.4); 1.31 (0.17–2.45)*  5.6 (0.3); 2.53 (1.55–3.51)**  6.5 (0.4); 3.48 (2.44–4.53)**    n (%); OR (95%CI) for adjusted model  PIMS  15 (20%); Reference  13 (25%); 1.26 (0.66–2.43)  24 (39%); 1.80 (1.04–3.10)*  11 (21%); 1.03 (0.52–2.03)  29 (32%); 1.55 (0.90–2.65)  21 (31%); 1.56 (0.88–2.76)  PPOS  23 (30%); Reference  25 (48%); 1.58 (1.02–2.45)*  38 (62%); 2.00 (1.34–2.97)**  22 (42%); 1.38 (0.86–2.21)  49 (53%); 1.76 (1.19–2.59)**  40 (59%); 1.85 (1.24–2.74)**  Pre 12-month admission  15 (17%); Reference  24 (36%); 1.98 (1.13–3.47)*  25 (30%); 1.68 (0.95–2.96)  18 (29%); 1.59 (0.87–2.91)*  36 (33%); 1.82 (1.06–3.11)*  44 (49%); 2.72 (1.63–4.53)**    Longitudinal outcomes of cluster membership n (%); HR (95%CI) for adjusted model  48-month any hospitalisationa  57 (65%); Reference  55 (83%); 1.32 (0.90–1.94)  71 (86%) 1.48 (1.03–2.12)*  48 (76%); 1.11 (0.75–1.65)  90 (82%); 1.30 (0.92–1.83)  70 (79%); 1.34 (0.94–1.92)  48-month mortalityb  11 (12%); Reference  17 (26%); 2.15 (0.98–4.74)  26 (31%); 2.66 (1.26–5.59)*  18 (29%); 2.32 (1.06–5.09)*  33 (30%); 2.55 (1.24–5.25)*  28 (32%); 2.77 (1.34–5.73)**    Longitudinal outcomes of number of conditions    <3 conditions, 231/501  ≥3 conditions, 270/501  48-month any hospitalisationa  168 (73%): Reference  223 (83%); 1.10 (0.88–1.36)  48-month mortalityb  38 (16%): Reference  95 (35%); 2.24 (1.52–3.29)**  CAD, coronary artery disease; CHF, congestive heart failure; TIA, transient ischaemic attack; PVD, peripheral vascular disease; DM, Type II diabetes mellitus; AF, atrial fibrillation; COPD, chronic obstructive pulmonary disease; RA, rheumatoid arthritis; SE, standard error; CI, confidence interval. The bold values highlight prevalences of conditions, not stats significance values. Missing data: Number for prescribed medications, PIMs, PPOs: 100/501 (20%); hospital admission: 3/501 (1%). Note: The prevalence of hypertension in Clusters 1 to 6: 85%, 77%, 81%, 83%, 83%, 93%, 84%. Base model to examine longitudinal outcomes includes gender, education, NZ deprivation index. aBase model plus pre-hospital admission, weighted survival time. bBase model plus 48-month admission. *P < 0.05; **P < 0.01. In assessing the stability of the cluster solution, dendrograms and scree plots from the random sample showed similar patterns and prevalence of conditions in clusters indicating the overall cluster solution had a high degree of stability. We identified six distinct clusters for Māori and non-Māori (Tables 1 and 2) and labelled as described in the method section, e.g. for Māori, Cluster 3 was labelled ‘arthritis’ as 84% of members had arthritis, and Cluster 4 was labelled ‘CVD, respiratory and mental health’ as CVD and asthma/chronic obstructive pulmonary disease (COPD) were highly prevalent and this cluster had the highest prevalence of depression across the clusters (38%). Note is made that the prevalence of some conditions in the ‘Well’ cluster is higher than might be expected from its given label (relative to prevalence in other clusters). Patterns of conditions and outcomes in Māori There were more women than men in all clusters except for the ‘CHF-AF’ cluster. Compared to the ‘Well’ cluster, members in other clusters had significantly more medications and higher odds of having a PPO (Table 1). Participants the ‘arthritis’ and ‘diabetes’ clusters had higher risk of hospitalisation than ‘Well’ cluster members. Overall, participants in ‘diabetes’ and ‘complex multi-morbidity’ clusters had the worst outcomes, 86% and 91% having at least one hospitalisation and 41% and 39% 48-month mortality. In contrast to the earlier analyses of <3 versus ≥3 conditions, these clusters showed greater differentiation of the risk of hospitalisation and mortality. Patterns of conditions and outcomes in non-Māori The clusters in non-Māori and Māori differed but were labelled as closely as possible to facilitate interpretation (Table 2). Compared to cluster ‘Well’, other clusters had significantly more medications and a higher risk of a PPO (except for cluster ‘cancer’). All clusters, except ‘CHF-AF’, had more than twice the risk of all-cause mortality compared with the ‘Well’ cluster members. Overall, those in cluster ‘depression and arthritis’ had the worst outcomes: 86% having at least one hospitalisation and 31% 48-month mortality. Diabetes did not cluster in one group and was more evenly spread through Clusters 2–6. Depression clustered with arthritis amongst non-Māori, and participants with stroke were mainly in Cluster 6 (‘Stroke’). Combinations of conditions, rather the count (<3 versus ≥3), showed greater differentiation of the risk of hospitalisation and mortality. Discussion This is the first study to our knowledge to follow hospitalisation and mortality associated with patterns of multi-morbidity in octogenarians. This analysis has highlighted that profiles of conditions carry stronger associations with cross-sectional and longitudinal outcomes than the sum of those conditions. In considering the six patterns of conditions in Māori and non-Māori, several clinical insights are evident. CVD rarely occurred in isolation; 96% of those with CVD had co-morbidities. Conditions that consistently clustered together in both cohorts were ‘CHF and AF’; in non-Māori, with osteoarthritis; in Māori, with depression and COPD. Outcomes for participants with atrial fibrillation (AF) differed whether they were in the ‘CHF-AF’ cluster (same mortality as the ‘Well’ group), or in the ‘Complex multi-morbidity’ cluster (2.58 times higher than the ‘Well’ group). Combinations and outcomes are ethnic specific. Attempts to understand multi-morbidity have examined varying populations [8, 19] but lack of standard definitions and processes hamper consistent conclusions [10]. As well as differentiating likelihood of poor outcomes, this analysis has highlighted the frequent combination of depression and respiratory disease for 25% of Māori octogenarians. These two conditions could be linked through lower activity from COPD contributing to depression, or lower COPD medication adherence because of depression [20]. Depression is associated with higher readmissions rates for COPD [21] and lower health-related QoL [22]. In contrast, depression clustered with arthritis in non-Māori and this cluster had particularly low quality prescribing and high mortality. We add to the NZ experience and international literature on mental and physical comorbidity showing that morbidity of mental and physical health predicted increased risk of hospitalisation and mortality. While there are calls to manage depression and other concurrent medical conditions [21] specific intervention development based on patterns of conditions and targeted trials may be needed to establish how best to improve outcomes. Mechanisms to mortality may be mediated through medication appropriateness which varied between clusters for both Māori and non-Māori. For both ethnic groups, PPOs were more common than PIMs and high mortality appeared to track more closely with high PPOs. PIMs rates have been associated with adverse drug events [23], but PPOs to date have not been linked to poorer outcomes. Prescribing patterns in Māori and non-Māori are explored further in another manuscript (under review). Of note, the criteria for quality of prescribing was based on current single disease guidelines. Further work is needed in understanding whether PIMs or PPOs are more relevant in predicting outcomes. The idea of ‘personalised medicine’ has gained credence internationally in recent years, frequently in the context of using molecular analysis to customisation healthcare in specific populations. The ageing population is heterogonous, and examining multi-morbidity patterns may prompt better ‘personalised care’. Smith et al. suggested that an integrated care management process comprising organisational (e.g. care co-ordination/management, regulatory interventions), pharmacologic and non-pharmacologic interventions (e.g. education, exercise therapy, cognitive behavioural therapy) focusing on specific area(s) is needed [24]. Further calls for precision in assessment of outcomes challenges current conceptualisations of management [25]. We add that a greater understanding of common patterns of conditions and their associated outcomes is needed. To complement ‘personalised care’, the healthcare system and clinical practice also needs to be cognisant of and embrace bi-cultural (e.g. NZ) and multi-cultural (e.g. USA, UK) older populations supporting healthcare professionals to be culturally competent in delivering care. We question whether multi-morbidity patterns may be consistent in differing populations. Very few studies have examined octogenarians, and those that have, have included hypertension as present in several clusters due to its high prevalence [8]. We excluded hypertension and thus potentially allowed underlying groupings to be identified. Some clusters identified here are similar to others in that coronary artery disease (CAD) and congestive heart failure (CHF) cluster together for Māori (‘complex multi-morbidity’), also observed in those aged 75+ in Sweden [26], and in the oldest old in Sweden dementia clustered with depression [26] as it did for both our cohorts. We did not identify cardio-metabolic clusters, common in other studies [8]. In this study when diabetes predominated, CVD was not co-prevalent in either ethnic group. Differences may be explained by different diagnostic groups included in studies and without a consistent platform populations cannot be compared. Larger studies with defined diagnostic groupings need to replicate this result. Further work should attempt consistency in multi-morbidity studies and develop specific treatment combinations related to common patterns. Other studies have used different techniques to identify clustering of conditions [8] (e.g. multiple correspondence analysis [27]; factor analysis [28, 29]; hierarchical cluster analysis by clustering of conditions [18]). Our clusters differed somewhat as we took a patient-centred approach and identified clusters of participants (rather than conditions) which could potentially lead to development of specific treatment regimes for specific groupings of co-existing conditions. In addition to different analytical techniques, different populations (age, ethnicity, geographical locations) and methods in ascertaining diagnosis (number and type of conditions) define other results [8, 27–30]. While the experience of NZ’s indigenous people is likely to be similar to those of indigenous peoples in other resource-rich countries with similar colonial histories, other indigenous peoples may have different histories and contemporary circumstances. Other ethnic diaspora ageing in a similar or new cultures/value system may also benefit from a separate examination of multi-morbidity. Strengths and limitations The strengths of this analysis are the certainty with which medications and diagnoses were identified, the inclusion of a representative group of the older indigenous population and the robustness of revalidated analysis. Limitations include the less stringent criteria (versus factor analysis) in deciding number of clusters in hierarchical clustering analysis [18]. Subjective criteria based on subject expertise were frequently employed in deciding number of clusters to ensure results are interpretable and meaningful, i.e. specificity to ensure manageability versus generalisability, and to optimise strategic attention to groupings of co-existing conditions. We pre-defined clinical criteria in deciding number of clusters. The low response rate (57%) [10] could be a potential source of bias (e.g. the frailer being less likely to participate). We included ‘ever’ diagnoses rather than ‘current’ diagnoses and this mainly affected the cancer group and perhaps the depression diagnosis. Missing data on outcome variables may lead to under estimation of associations between outcomes and clusters, a Type II error. This is less likely for hospitalisations and mortality with a National Health Index number, but interview data for medications was not universally available. These findings will better inform the implementation of health programming and policies for healthcare delivery for the aged through emphasis on specific patterns rather than individual diseases in educational programmes, stimulation of specific trial research on outcomes related to different treatment regimens; and through raising public awareness of the complexity of multi-morbidity. Conclusions We have shown that certain disease combinations put older people at particular risk of poor outcome. These findings support personalised care for different ethnic groups based on differences between the patterns of multi-morbidity and observed associated outcomes. Further work will examine treatment patterns and whether and how they associate with outcomes within these combinations of multi-morbidity. Key points Clusters of conditions provide better differentiation of hospitalisation and mortality risk than count of conditions. Different ethnic groups have distinctive shared conditions with Cardiovascular Disease (CVD). Better ‘personalised care’ requires examining patterns of multi-morbidity. Supplementary data Supplementary data mentioned in the text are available to subscribers in Age and Ageing online. Funding This work was supported by the Health Research Council of New Zealand (HRC 09/068B; UoA ref: 3624940) and Ministry of Health New Zealand (MOH ref: 345426/00; UoA ref 3703221) which funded the project management and data collection work; Ngā Pae o te Māramatanga (UoA ref: 3624946) which funded the Māori engagement and project management; New Zealand Heart Foundation project grant for investigating cardiac markers (UoA Ref: 3625921) and a Heart Foundation Research Fellowship (UoA ref: 3702288). We thank the sponsors. Conflict of interest None. Acknowledgements We wish to acknowledge the participants, their families and whanau for supporting the study. We are extremely grateful to Dr. Lorna Dyall (PhD) and Dr. Mere Kepa (EdD) for the significant contributions in drafting this manuscript who are now enjoying their retirement days. Their inputs on interpretation of data from Māori worldview are instrumental. We thank Te RōpuKaitiaki o ngā tikanga Māori for their guidance. We acknowledge the community partners in LiLACS NZ who engaged with participants and collected data (Western Bay of Plenty Primary Health Organisation, Ngā Matāpuna Oranga Kaupapa Māori Primary Health Organisation, Te Korowai Aroha Trust, Te Rūnanga o Ngati Pikiao, Rotorua Area Primary Health Services, Ngati Awa Research & Archives Trust, Te Rūnanga o Ngati Irapuaia and Te Whanau a Apanui Community Health Centre). References 1 Barnett K, Mercer SW, Norbury M et al.  . Epidemiology of multimorbidity and implications for health care, research, and medical education: a cross-sectional study. Lancet  2012; 380: 37– 43. Google Scholar CrossRef Search ADS PubMed  2 Fortin M, Hudon C, Haggerty J et al.  . Prevalence estimates of multimorbidity: a comparative study of two sources. BMC Health Serv Res  2010; 10: 111. Google Scholar CrossRef Search ADS PubMed  3 Aarts S, den Akker M.v., Bosma H et al.  . The effect of multimorbidity on health related functioning: Temporary or persistent? Results from a longitudinal cohort study. J Psychosom Res  2012; 73: 211– 7. Google Scholar CrossRef Search ADS PubMed  4 Aarts S, van den Akker M, Hajema KJ et al.  . Multimorbidity and its relation to subjective memory complaints in a large general population of older adults. Int Psychogeriatr  2011; 23: 616– 24. Google Scholar CrossRef Search ADS PubMed  5 Fortin M, Dubois M-F, Hudon C et al.  . Multimorbidity and quality of life: a closer look. Health Qual Life Outcomes  2007; 5: 52. Google Scholar CrossRef Search ADS PubMed  6 van den Bussche H, Schon G, Kolonko T et al.  . Patterns of ambulatory medical care utilization in elderly patients with special reference to chronic diseases and multimorbidity—results from a claims data based observational study in Germany. BMC Geriatr  2011; 11: 54. Google Scholar CrossRef Search ADS PubMed  7 Fortin M, Stewart M, Poitras M-E et al.  . A systematic review of prevalence studies on multimorbidity: toward a more uniform methodology. Ann Fam Med  2012; 10: 142– 51. Google Scholar CrossRef Search ADS PubMed  8 Violan C, Foguet-Boreu Q, Flores-Mateo G et al.  . Prevalence, determinants and patterns of multimorbidity in Primary Care: a systematic review of observational studies. PLOS one  2014; 9: e102149. Google Scholar CrossRef Search ADS PubMed  9 Kerse N, Teh R, Moyes S et al.  . Cohort profile: Te Puawaitanga o Nga Tapuwae Kia Ora Tonu, Life and Living in Advanced Age: a Cohort Study in New Zealand (LiLACS NZ). Int J Epidemiol  2015; 44: 1823– 32. Google Scholar CrossRef Search ADS PubMed  10 Dyall L, Kepa M, Hayman K et al.  . Engagement and recruitment of Māori and non-Māori people of advanced age to LiLACS NZ. Aust N Z J Public Health  2013; 37: 124– 31. Google Scholar CrossRef Search ADS PubMed  11 Teh R, Doughty R, Connolly M et al.  . Agreement between self-reports and medical records of cardiovascular disease in octogenarians. J Clin Epidemiol  2013; 66: 1135– 43. Google Scholar CrossRef Search ADS PubMed  12 World Health Organisation. World Health Organisation Collaborating Centre for Drug Statistics Methodology. [cited 2011; Available from: http://www.whocc.no/. 13 Gallagher P, Ryan C, Byrne S et al.  . STOPP (Screening Tool of Older Person’s Prescriptions) and START (Screening Tool to Alert doctors to Right Treatment). Consensus validation. Int J Clin Pharmacol Ther  2008; 46: 72– 83. Google Scholar CrossRef Search ADS PubMed  14 Teh R, Kerse N, Kepa M et al.  . Self-rated health, health related behaviours and medical conditions of Māori and non-Māori in advanced age: LiLACS NZ. N Z Med J  2014; 127: 13– 29. Google Scholar PubMed  15 Hill S, Sarfati D, Blakely T et al.  . Survival disparities in Indigenous and non-Indigenous New Zealanders with colon cancer: the role of patient comorbidity, treatment and health service factors. J Epidemiol Community Health  2010; 64: 117– 23. Google Scholar CrossRef Search ADS PubMed  16 Chan WC, Wright C, Riddell T et al.  . Ethnic and socioeconomic disparities in the prevalence of cardiovascular disease in New Zealand. NZMJ  2008; 121: 11– 20. 17 Pihama L. A conversation about kaupapa Māori theory and research. In: Hutchings J, Potter H, Taupo K, eds. New Zealand Council for Educational Research: Challenges of Kaupapa Māori Research . Wellington: Pipitea marae, 2011; pp. 49– 55. 18 Cornell JE, Pugh JA, Williams JW Jr et al.  . Multimorbidity clusters: clustering binary data from multimorbidity clusters: clustering binary data from a large administrative medical database. Appl Multivar Res  2009; 12: 163– 82. Google Scholar CrossRef Search ADS   19 Sinnige J, Korevaar JC, Westert GP et al.  . Multimorbidity patterns in a primary care population aged 55 years and over. Fam Pract  2015; 32: 505– 13. Google Scholar CrossRef Search ADS PubMed  20 Bender BG. Nonadherence in chronic obstructive pulmonary disease patients: what do we know and what should we do next? Curr Opin Pulm Med  2014; 20: 132– 7. Google Scholar CrossRef Search ADS PubMed  21 Pooler A, Beech R. Examining the relationship between anxiety and depression and exacerbations of COPD which result in hospital admission: a systematic review. Int J Copd  2014; 9: 315– 30. Google Scholar CrossRef Search ADS   22 Sundh J, Johansson G, Larsson K et al.  . Comorbidity and health-related quality of life in patients with severe chronic obstructive pulmonary disease attending Swedish secondary care units. Int J Copd  2015; 10: 173– 83. Google Scholar CrossRef Search ADS   23 Hamilton H, Gallagher P, Ryan C et al.  . Potentially inappropriate medications defined by STOPP criteria and the risk of adverse drug events in older hospitalized patients. Arch Int Med  2011; 171: 1013– 9. Google Scholar CrossRef Search ADS   24 Smith SM, Soubhi H, Fortin M et al.  . Managing patients with multimorbidity: systematic review of interventions in primary care and community settings. Br Med J  2012; 345: e5205. Google Scholar CrossRef Search ADS   25 Bierman AS, Tinetti ME. Precision medicine to precision care: managing multimorbidity. Lancet  2016; 388: 2721– 3. Google Scholar CrossRef Search ADS PubMed  26 Marengoni A, Rizzuto D, Wang HX et al.  . Patterns of chronic multimorbidity in the elderly population. J Am Geriatr Soc  2009; 57: 225– 30. Google Scholar CrossRef Search ADS PubMed  27 Garcia-Olmos L, Salvador CH, Alberquilla A et al.  . Comorbidity patterns in patients with chronic diseases in general practice. PLoS ONE  2012; 7: e32141. Google Scholar CrossRef Search ADS PubMed  28 Schäfer I, von Leitner E-C, Schön G et al.  . Multimorbidity patterns in the elderly: a new approach of disease clustering identifies complex interrelations between chronic conditions. PLoS ONE  2010; 5: e15941. Google Scholar CrossRef Search ADS PubMed  29 Kirchberger I, Meisinger C, Heier M et al.  . Patterns of multimorbidity in the aged population. Results from the KORA-Age study. PLoS ONE  2012; 7: e30556. Google Scholar CrossRef Search ADS PubMed  30 Steinman MA, Lee SJ, Boscardin WJ et al.  . Patterns of multimorbidity in elderly veterans. J Am Geriatr Soc  2012; 60: 1872– 80. Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2017. Published by Oxford University Press on behalf of the British Geriatrics Society.All rights reserved. For permissions, please email: journals.permissions@oup.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Age and Ageing Oxford University Press

Patterns of multi-morbidity and prediction of hospitalisation and all-cause mortality in advanced age

Loading next page...
 
/lp/ou_press/patterns-of-multi-morbidity-and-prediction-of-hospitalisation-and-all-2ztnyQ7UTJ
Publisher
Oxford University Press
Copyright
© The Author(s) 2017. Published by Oxford University Press on behalf of the British Geriatrics Society.All rights reserved. For permissions, please email: journals.permissions@oup.com
ISSN
0002-0729
eISSN
1468-2834
D.O.I.
10.1093/ageing/afx184
Publisher site
See Article on Publisher Site

Abstract

Abstract Background multi-morbidity is associated with poor outcomes and increased healthcare utilisation. We aim to identify multi-morbidity patterns and associations with potentially inappropriate prescribing (PIP), subsequent hospitalisation and mortality in octogenarians. Methods life and Living in Advanced Age; a Cohort Study in New Zealand (LiLACS NZ) examined health outcomes of 421 Māori (indigenous to New Zealand), aged 80–90 and 516 non-Māori, aged 85 years in 2010. Presence of 14 chronic conditions was ascertained from self-report, general practice and hospitalisation records and physical assessments. Agglomerative hierarchical cluster analysis identified clusters of participants with co-existing conditions. Multivariate regression models examined the associations between clusters and PIP, 48-month hospitalisations and mortality. Results six clusters were identified for Māori and non-Māori, respectively. The associations between clusters and outcomes differed between Māori and non-Māori. In Māori, those in the complex multi-morbidity cluster had the highest prevalence of inappropriately prescribed medications and in cluster ‘diabetes’ (20% of sample) had higher risk of hospitalisation and mortality at 48-month follow-up. In non-Māori, those in the ‘depression-arthritis’ (17% of the sample) cluster had both highest prevalence of inappropriate medications and risk of hospitalisation and mortality. Conclusions in octogenarians, hospitalisation and mortality are better predicted by profiles of clusters of conditions rather than the presence or absence of a specific condition. Further research is required to determine if the cluster approach can be used to target patients to optimise resource allocation and improve outcomes. comorbidity, hospitalisation, aged, indigenous population, older people Introduction Multi-morbidity increases with age [1, 2] being present in up to 82% of individuals aged over 85 years [1]. Multi-morbidity is associated with: rapid functional decline amongst older people [3]; memory problems [4]; lower quality of life (QoL) [5], and is a stronger predictor of healthcare utilisation than age or gender [6]. The ageing of the population, with those aged 85+ being both the fastest growing group and the most likely to have multi-morbidity, will mean understanding and managing multi-morbidity in the oldest old is increasingly important. Studies of multi-morbidity lack consistency of methods [7]. Patterns of diseases may be more useful than counting number of conditions. While patterns of multi-morbidity have been identified [8], studies to date seldom link outcomes to patterns, and none have examined octogenarians or indigenous older populations. New understandings of outcomes related to specific disease groupings may lead to new strategies in improving patient-centred care and outcomes. We aim to identify specific patterns of conditions of octogenarians living in NZ, contrasting two ethnic groups, and report associated prescribing appropriateness and subsequent hospitalisations and mortality over 4 years. Methods Te Puāwaitanga O Nga Tapuwae Kia ora Tonu: Life and Living in Advanced Age; a Cohort Study in NZ (LiLACS NZ) is a cohort study of Māori (indigenous people in New Zealand) and non-Māori octogenarians [9]. The study recruited a population based sample of 421 Māori aged 80–90 years and 516 non-Māori aged 85 years between March 2010 and April 2011 from a regionally defined area in the North Island, NZ. Detailed recruitment strategies are previously described [10]. The study was approved by the Northern X Regional Ethics Committee (NXT/09/09/088; NXT/10/12/127). Written informed consent was obtained from all participants. Data collection Socio-economic, health, and physical status were established using standardised techniques (comprehensive face-to-face interviews and physical assessments). For participants who scored <73 in the Modified Mini Mental Status Examination, a proxy was identified and completed the interview (details in Supplementary material). Diagnoses Fourteen pre-specified medical conditions prevalent in older adults were identified from self-report (‘Have you ever been told by a doctor that you have had [condition]?’), General Practice (GP) records (list of conditions), hospitalisation records, physical assessments (Figure 1). Only participants having a record of presence/absence of respective conditions in one of the five sources were included in analyses [11]. Figure 1. View largeDownload slide Data sources for participants in LiLACS NZ. LILACS NZ, Life and Living in Advanced Age; a Cohort Study in New Zealand; GP, general practitioner; NZHIS, New Zealand Health Identification System aQuestionnaire: 5 participants did not complete questionnaire. bGP medical records: No consent, n = 41. Consented but review was not completed (n = 59), Consented but participants changed their mind (n = 24), Consented but GP refused to give out information (n = 2). cNZHIS: 60 participants refused consent. dPhysical assessments: No consent, n = 306. Of those who consented (n = 630), 624 had blood pressure measurements, six did not have BP measurements (participants refused, n = 3; unable to do, n = 2, participant out of town); 615 had an ECG read by cardiologist, 15 ECGs were not received (participants refused ECG n = 4; unable to physically co-operate, n = 3; unsuitable environment n = 2; other reasons n = 6). eBlood test: No consent, n = 305. Consented but did not gave sample, n = 76. Of those who consented (n = 556), 546 had glucose analysis, 10 did not have glucose analysis due to laboratory oversight. Figure 1. View largeDownload slide Data sources for participants in LiLACS NZ. LILACS NZ, Life and Living in Advanced Age; a Cohort Study in New Zealand; GP, general practitioner; NZHIS, New Zealand Health Identification System aQuestionnaire: 5 participants did not complete questionnaire. bGP medical records: No consent, n = 41. Consented but review was not completed (n = 59), Consented but participants changed their mind (n = 24), Consented but GP refused to give out information (n = 2). cNZHIS: 60 participants refused consent. dPhysical assessments: No consent, n = 306. Of those who consented (n = 630), 624 had blood pressure measurements, six did not have BP measurements (participants refused, n = 3; unable to do, n = 2, participant out of town); 615 had an ECG read by cardiologist, 15 ECGs were not received (participants refused ECG n = 4; unable to physically co-operate, n = 3; unsuitable environment n = 2; other reasons n = 6). eBlood test: No consent, n = 305. Consented but did not gave sample, n = 76. Of those who consented (n = 556), 546 had glucose analysis, 10 did not have glucose analysis due to laboratory oversight. Outcomes Medication use was established by direct observation of medications (prescribed and non-prescribed), participants were asked how it was taken and were coded with WHO-ATC codes [12]. Prescribing quality was assessed using STOPP/START criteria [13]. STOPP identifies potentially inappropriate medications (PIMs); START lists potential prescribing omissions (PPOs). Hospitalisations and mortality were ascertained from NZ Ministry of Health data based on individually matched National Health Identification number for 48 months post-enrolment. Statistical analysis We chose to examine Māori and non-Māori separately because we know that prevalence of diseases [14] and outcomes from these conditions differ between Māori and non-Māori [15, 16], and different ethnic groups may need different interventions strategies [17]. Patient-centred approaches necessitate patient-centred rather than condition-centred analyses. Agglomerative hierarchical cluster analysis was used to identify clinically relevant subgroups based on groupings of co-existing conditions on all participants with complete data. An example of this method is species taxonomy. For the clustering algorithm Ward’s method was used. This sorts individuals into clusters containing a similar profile of conditions [18]. The number of clusters was guided by a combination of methods including visualising the dendrograms and scree plots. Determining the best number of clusters is a balance between manageability and strategic attention. We adapted Cornell’s pre-defined criteria [18], meaning the cluster solution was based on informed judgements and consensus of health researchers (geriatricians, cardiologist, general practitioner, biostatisticians, epidemiologist). Stability of the cluster solution was assessed by repeating the process on a randomly selecting 50% of the Māori and non-Māori sample. Cluster labelling was based on the most prevalent conditions within the cluster and highest representation of the condition across clusters. Descriptive statistics are presented for each of the clusters. Multivariate regression models (linear or log-binomial regression) examined associations between identified clusters and PIMs, PPOs, and pre 12-month hospitalisation adjusting for age (Māori only), gender, education status and deprivation index. Cox-proportional hazards models were used to determine hazard ratio (HR) for 48-month hospitalisation and all-cause mortality. We used the cluster identified as ‘healthiest’ (fewer conditions or lower prevalence of all conditions versus other clusters) as the reference group in regression analyses. Separate analyses (adjusted for similar confounders) were completed based on median number of conditions. Statistical analyses were performed with IBM SPSS v23. A P-value < 0.05, two-tailed, was considered statistically significant. Results Participant characteristics In the study, 888/937 participants were included in analyses (387/421 Māori; 501/516 non-Māori) (Figure 1). The proportion of participants living in residential care was similar between those included and excluded (8% vs. 13%, P = 0.176), as with functional status (NEADL score), median(IQR) 19(4) vs. 17(5), P = 0.589. Among the 387 Māori, 221 (57%) were women; mean age was 82.7 years (SD 2.8). All the 501 non-Māori were 85 years old at entry; 268 (54%) were women. The median(IQR) for number of conditions overall was 3(3) for both Māori and non-Māori. Inappropriate medication was identified in 168 (66%) Māori and 250 (62%) non-Māori. Over 48-month follow-up, all-cause hospitalisations occurred in 302 (80%) Māori and 391 (79%) non-Māori; 121 (31%) Māori and 133 (27%) non-Māori died. For Māori, those with ≥3 conditions had higher risk of any hospitalisation (P < 0.01) and all-cause mortality (P < 0.001) at 48-month follow-up than those with <3 conditions (Table 1). Non-Māori with ≥3 conditions had a higher risk of all-cause mortality (P < 0.001) at 48-month follow-up (Table 2). The risk of any hospitalisation did not differ between those with less than three conditions and those having three or more. Table 1. Prevalence, patterns of conditions and outcomes for Māori   Cluster 1, 73/387 (19%)  Cluster 2, 59/387 (15%)  Cluster 3, 45/387 (12%)  Cluster 4, 98/387 (25%)  Cluster 5, 79/387 (20%)  Cluster 6, 33/387 (9%)  ‘Well’ (lower prevalence of conditions)  CHF and AF  Arthritis  CVD, respiratory and mental health  Diabetes  Complex multi-morbidity (high prevalence of most conditions)    Mean (SD)  Age  82.5 (2.9)  83.2 (3.0)  83.3 (2.5)  82.8 (2.7)  82.4 (2.5)  82.4 (2.8)  Number of conditions  1.1 (1.3)  3.7 (1.4)  2.7 (1.2)  3.4 (1.9)  3.8 (1.7)  6.5 (1.4)    n (%)  Sex, male  30 (41)  31 (53)  13 (29)  41 (42)  37 (47)  14 (42)  Deprivation Index     Low  8 (11)  11 (19)  9 (20)  13 (13)  11 (14)  5 (15)   Medium  25 (34)  18 (30)  9 (20)  23 (24)  21 (27)  6 (18)   High  40 (55)  30 (51)  27 (60)  62 (63)  47 (59)  22 (67)  CAD/PVD  1 (1)  41 (69)  14 (31)  71 (72)  47 (59)  31 (94)  CHF  1 (1)  43 (73)  4 (9)  15 (15)  39 (49)  22 (67)  Stroke/TIA  4 (5)  13 (22)  8 (18)  43 (44)  20 (25)  7 (21)  Any AF  3 (4)  52 (88)  5 (11)  7 (7)  30 (38)  21 (64)  DM  2 (3)  0 (0)  9 (20)  17 (17)  75 (95)  15 (45)  Asthma or COPD  7 (10)  14 (24)  1 (2)  62 (63)  21 (27)  30 (91)  Osteoporosis  0 (0)  4 (7)  21 (47)  15 (15)  6 (8)  16 (48)  Osteoarthritis  5 (7)  9 (15)  38 (84)  23 (23)  4 (5)  31 (94)  RA  13 (18)  2 (3)  7 (16)  10 (10)  13 (16)  24 (73)  Dementia  12 (16)  11 (19)  6 (13)  23 (23)  11 (14)  0 (0)  Depression  12 (16)  17 (29)  6 (13)  37 (38)  15 (19)  9 (27)  Thyroid disease  3 (4)  0 (0)  0 (0)  4 (4)  4 (5)  3 (9)  Non-skin cancer  13 (18)  11 (19)  1 (2)  8 (8)  17 (22)  6 (18)  Melanoma  2 (3)  0 (0)  0 (0)  1 (1)  0 (0)  0 (0)    Cross-sectional associations with cluster membership Mean (SE); Beta (95%CI) for adjusted model  Prescribed medications  2.4 (0.4); Reference  5.6 (0.5); 3.25 (2.03–4.47)**  3.6 (0.5); 1.25 (0.04–2.46)*  5.1 (0.4); 2.68 (1.65–3.71)**  6.2 (0.4); 3.84 (2.76–4.92)**  7.9 (0.5); 5.54 (4.30–6.79)**    n (%); OR (95%CI) for adjusted model  PIMS  6 (11%); Reference  15 (45%); 4.05 (1.72–9.56)**  5 (15%); 1.45 (0.47–4.45)  8 (14%); 1.20 (0.43–3.31)  12 (25%); 2.05 (0.83–5.06)  15 (52%); 4.51 (1.94–10.50)**  PPOS  11 (21%); Reference  16 (49%); 2.57 (1.34–4.96)**  22 (65%); 3.56 (1.97–6.45)**  40 (69%); 3.52 (1.99–6.22)**  36 (74%); 3.87 (2.19–6.83)**  25 (86%); 4.32 (2.45–7.61)**  Pre 12-month admission  10 (14%); Reference  26 (45%); 3.23 (1.63–6.41)**  11 (25%); 2.20 (1.00–5.00)  37 (39%); 3.00 (1.56–5.81)**  35 (45%); 3.46 (1.81–6.65)**  21 (66%); 4.82 (2.48–9.38)**    Longitudinal outcomes of cluster membership n (%); HR (95%CI) for adjusted model  48-month any hospitalisationa  48 (67%); Reference  48 (83%); 1.41 (0.93–2.15)  36 (82%): 1.77 (1.12–2.79)*  74 (78%): 1.29 (0.88–1.89)  67 (86%); 1.51 (1.02–2.22)*  29 (91%); 1.70 (1.05–2.77)*  48-month mortalityb  13 (18%): Reference  16 (27%); 1.47 (0.68–3.15)  14 (31%); 1.56 (0.70–3.50)  33 (34%); 2.02 (1.03–3.95)*  32 (41%); 2.58 (1.32–5.05)**  13 (39%); 2.58 (1.17–5.66)*    Longitudinal outcomes of number of conditions    <3 conditions, 147/387  ≥3 conditions, 240/387  48-month any hospitalisationa  100 (70%); Reference  202 (85%); 1.54 (1.19–1.99)**  48-month mortalityb  27 (18%); Reference  94 (39%): 2.46 (1.58–3.83)**    Cluster 1, 73/387 (19%)  Cluster 2, 59/387 (15%)  Cluster 3, 45/387 (12%)  Cluster 4, 98/387 (25%)  Cluster 5, 79/387 (20%)  Cluster 6, 33/387 (9%)  ‘Well’ (lower prevalence of conditions)  CHF and AF  Arthritis  CVD, respiratory and mental health  Diabetes  Complex multi-morbidity (high prevalence of most conditions)    Mean (SD)  Age  82.5 (2.9)  83.2 (3.0)  83.3 (2.5)  82.8 (2.7)  82.4 (2.5)  82.4 (2.8)  Number of conditions  1.1 (1.3)  3.7 (1.4)  2.7 (1.2)  3.4 (1.9)  3.8 (1.7)  6.5 (1.4)    n (%)  Sex, male  30 (41)  31 (53)  13 (29)  41 (42)  37 (47)  14 (42)  Deprivation Index     Low  8 (11)  11 (19)  9 (20)  13 (13)  11 (14)  5 (15)   Medium  25 (34)  18 (30)  9 (20)  23 (24)  21 (27)  6 (18)   High  40 (55)  30 (51)  27 (60)  62 (63)  47 (59)  22 (67)  CAD/PVD  1 (1)  41 (69)  14 (31)  71 (72)  47 (59)  31 (94)  CHF  1 (1)  43 (73)  4 (9)  15 (15)  39 (49)  22 (67)  Stroke/TIA  4 (5)  13 (22)  8 (18)  43 (44)  20 (25)  7 (21)  Any AF  3 (4)  52 (88)  5 (11)  7 (7)  30 (38)  21 (64)  DM  2 (3)  0 (0)  9 (20)  17 (17)  75 (95)  15 (45)  Asthma or COPD  7 (10)  14 (24)  1 (2)  62 (63)  21 (27)  30 (91)  Osteoporosis  0 (0)  4 (7)  21 (47)  15 (15)  6 (8)  16 (48)  Osteoarthritis  5 (7)  9 (15)  38 (84)  23 (23)  4 (5)  31 (94)  RA  13 (18)  2 (3)  7 (16)  10 (10)  13 (16)  24 (73)  Dementia  12 (16)  11 (19)  6 (13)  23 (23)  11 (14)  0 (0)  Depression  12 (16)  17 (29)  6 (13)  37 (38)  15 (19)  9 (27)  Thyroid disease  3 (4)  0 (0)  0 (0)  4 (4)  4 (5)  3 (9)  Non-skin cancer  13 (18)  11 (19)  1 (2)  8 (8)  17 (22)  6 (18)  Melanoma  2 (3)  0 (0)  0 (0)  1 (1)  0 (0)  0 (0)    Cross-sectional associations with cluster membership Mean (SE); Beta (95%CI) for adjusted model  Prescribed medications  2.4 (0.4); Reference  5.6 (0.5); 3.25 (2.03–4.47)**  3.6 (0.5); 1.25 (0.04–2.46)*  5.1 (0.4); 2.68 (1.65–3.71)**  6.2 (0.4); 3.84 (2.76–4.92)**  7.9 (0.5); 5.54 (4.30–6.79)**    n (%); OR (95%CI) for adjusted model  PIMS  6 (11%); Reference  15 (45%); 4.05 (1.72–9.56)**  5 (15%); 1.45 (0.47–4.45)  8 (14%); 1.20 (0.43–3.31)  12 (25%); 2.05 (0.83–5.06)  15 (52%); 4.51 (1.94–10.50)**  PPOS  11 (21%); Reference  16 (49%); 2.57 (1.34–4.96)**  22 (65%); 3.56 (1.97–6.45)**  40 (69%); 3.52 (1.99–6.22)**  36 (74%); 3.87 (2.19–6.83)**  25 (86%); 4.32 (2.45–7.61)**  Pre 12-month admission  10 (14%); Reference  26 (45%); 3.23 (1.63–6.41)**  11 (25%); 2.20 (1.00–5.00)  37 (39%); 3.00 (1.56–5.81)**  35 (45%); 3.46 (1.81–6.65)**  21 (66%); 4.82 (2.48–9.38)**    Longitudinal outcomes of cluster membership n (%); HR (95%CI) for adjusted model  48-month any hospitalisationa  48 (67%); Reference  48 (83%); 1.41 (0.93–2.15)  36 (82%): 1.77 (1.12–2.79)*  74 (78%): 1.29 (0.88–1.89)  67 (86%); 1.51 (1.02–2.22)*  29 (91%); 1.70 (1.05–2.77)*  48-month mortalityb  13 (18%): Reference  16 (27%); 1.47 (0.68–3.15)  14 (31%); 1.56 (0.70–3.50)  33 (34%); 2.02 (1.03–3.95)*  32 (41%); 2.58 (1.32–5.05)**  13 (39%); 2.58 (1.17–5.66)*    Longitudinal outcomes of number of conditions    <3 conditions, 147/387  ≥3 conditions, 240/387  48-month any hospitalisationa  100 (70%); Reference  202 (85%); 1.54 (1.19–1.99)**  48-month mortalityb  27 (18%); Reference  94 (39%): 2.46 (1.58–3.83)**  CAD, coronary artery disease; CHF, congestive heart failure; TIA, transient ischaemic attack; PVD, peripheral vascular disease; DM, Type II diabetes mellitus; AF, atrial fibrillation; COPD, chronic obstructive pulmonary disease; RA, rheumatoid arthritis; CVD, cardiovascular disease; SE, standard error; CI, confidence interval. Missing data: Number for prescribed medications, PIMs, PPOs: 130/387 (34%); hospital admission: 8/387 (2%). The bold values highlight prevalences of conditions, not stats significance values. Note: The prevalence of hypertension in Clusters 1 to 6: 87%, 97%, 82%, 87%, 75%, 94%. Base model to examine longitudinal outcomes includes age, gender, education, NZ deprivation index. aBase model plus pre-hospital admission, weighted survival time; bBase model plus 48-month admission. *P < 0.05; **P < 0.01. Table 2. Prevalence and patterns of conditions and outcomes for non-Māori   Cluster 1, 89/501 (18%)  Cluster 2, 66/501 (13%)  Cluster 3, 83/501 (17%)  Cluster 4, 63/501 (13%)  Cluster 5, 111/501 (22%)  Cluster 6, 89/501 (18%)  ‘Well’ (lower prevalence of conditions)  CHF and AF  Depression and arthritis  Cancer  Respiratory and diabetes  Stroke    Mean (SD)  Age  All non-Māori participants born in the same year, 84.6 (0.6)  Number of conditions  0.9 (1.1)  3.2 (1.4)  3.3 (1.6)  2.9 (1.3)  3.0 (1.7)  4.0 (2.1)    n (%)  Sex, male  34 (38)  41 (62)  37 (45)  36 (57)  51 (46)  34 (38)  Deprivation index     Low  31 (35)  18 (27)  11 (13)  17 (27)  32 (21)  23 (26)   Medium  37 (42)  27 (41)  33 (40)  22 (35)  51 (46)  34 (38)   High  21 (24)  21 (32)  39 (47)  24 (38)  37 (33)  32 (36)  CAD/PVD  3 (3)  48 (73)  27 (33)  30 (48)  69 (62)  58 (65)  CHF  3 (3)  30 (45)  7 (8)  10 (16)  30 (27)  27 (30)  Stroke/TIA  1 (1)  5 (8)  17 (20)  3 (5)  26 (23)  84 (94)  Any AF  3 (3)  50 (76)  5 (6)  5 (8)  15 (14)  26 (29)  DM  0 (0)  13 (20)  16 (19)  14 (22)  30 (27)  6 (7)  Asthma or COPD  0 (0)  7 (11)  19 (23)  12 (19)  80 (72)  16 (18)  Osteoporosis  27 (30)  2 (3)  24 (29)  9 (14)  18 (16)  25 (28)  Osteoarthritis  27 (30)  30 (45)  47 (57)  15 (24)  34 (31)  46 (52)  RA  0 (0)  3 (5)  34 (41)  0 (0)  12 (11)  2 (2)  Dementia  10 (11)  8 (12)  16 (19)  1 (2)  7 (6)  5 (6)  Depression  4 (4)  10 (15)  61 (73)  10 (16)  1 (1)  37 (42)  Thyroid disease  4 (4)  3 (5)  4 (5)  6 (10)  7 (6)  13 (15)  Non-skin cancer  0 (0)  5 (8)  6 (7)  59 (94)  4 (4)  16 (18)  Melanoma  4 (4)  3 (5)  5 (6)  11 (17)  6 (5)  2 (2)    Cross-sectional associations with cluster membership Mean (SE); Beta (95%CI) for adjusted model  Prescribed medications, n  3.1 (0.4); Reference  5.5 (0.4); 2.50 (1.37–3.63)**  5.6 (0.4); 2.58 (1.48–3.68)**  4.4 (0.4); 1.31 (0.17–2.45)*  5.6 (0.3); 2.53 (1.55–3.51)**  6.5 (0.4); 3.48 (2.44–4.53)**    n (%); OR (95%CI) for adjusted model  PIMS  15 (20%); Reference  13 (25%); 1.26 (0.66–2.43)  24 (39%); 1.80 (1.04–3.10)*  11 (21%); 1.03 (0.52–2.03)  29 (32%); 1.55 (0.90–2.65)  21 (31%); 1.56 (0.88–2.76)  PPOS  23 (30%); Reference  25 (48%); 1.58 (1.02–2.45)*  38 (62%); 2.00 (1.34–2.97)**  22 (42%); 1.38 (0.86–2.21)  49 (53%); 1.76 (1.19–2.59)**  40 (59%); 1.85 (1.24–2.74)**  Pre 12-month admission  15 (17%); Reference  24 (36%); 1.98 (1.13–3.47)*  25 (30%); 1.68 (0.95–2.96)  18 (29%); 1.59 (0.87–2.91)*  36 (33%); 1.82 (1.06–3.11)*  44 (49%); 2.72 (1.63–4.53)**    Longitudinal outcomes of cluster membership n (%); HR (95%CI) for adjusted model  48-month any hospitalisationa  57 (65%); Reference  55 (83%); 1.32 (0.90–1.94)  71 (86%) 1.48 (1.03–2.12)*  48 (76%); 1.11 (0.75–1.65)  90 (82%); 1.30 (0.92–1.83)  70 (79%); 1.34 (0.94–1.92)  48-month mortalityb  11 (12%); Reference  17 (26%); 2.15 (0.98–4.74)  26 (31%); 2.66 (1.26–5.59)*  18 (29%); 2.32 (1.06–5.09)*  33 (30%); 2.55 (1.24–5.25)*  28 (32%); 2.77 (1.34–5.73)**    Longitudinal outcomes of number of conditions    <3 conditions, 231/501  ≥3 conditions, 270/501  48-month any hospitalisationa  168 (73%): Reference  223 (83%); 1.10 (0.88–1.36)  48-month mortalityb  38 (16%): Reference  95 (35%); 2.24 (1.52–3.29)**    Cluster 1, 89/501 (18%)  Cluster 2, 66/501 (13%)  Cluster 3, 83/501 (17%)  Cluster 4, 63/501 (13%)  Cluster 5, 111/501 (22%)  Cluster 6, 89/501 (18%)  ‘Well’ (lower prevalence of conditions)  CHF and AF  Depression and arthritis  Cancer  Respiratory and diabetes  Stroke    Mean (SD)  Age  All non-Māori participants born in the same year, 84.6 (0.6)  Number of conditions  0.9 (1.1)  3.2 (1.4)  3.3 (1.6)  2.9 (1.3)  3.0 (1.7)  4.0 (2.1)    n (%)  Sex, male  34 (38)  41 (62)  37 (45)  36 (57)  51 (46)  34 (38)  Deprivation index     Low  31 (35)  18 (27)  11 (13)  17 (27)  32 (21)  23 (26)   Medium  37 (42)  27 (41)  33 (40)  22 (35)  51 (46)  34 (38)   High  21 (24)  21 (32)  39 (47)  24 (38)  37 (33)  32 (36)  CAD/PVD  3 (3)  48 (73)  27 (33)  30 (48)  69 (62)  58 (65)  CHF  3 (3)  30 (45)  7 (8)  10 (16)  30 (27)  27 (30)  Stroke/TIA  1 (1)  5 (8)  17 (20)  3 (5)  26 (23)  84 (94)  Any AF  3 (3)  50 (76)  5 (6)  5 (8)  15 (14)  26 (29)  DM  0 (0)  13 (20)  16 (19)  14 (22)  30 (27)  6 (7)  Asthma or COPD  0 (0)  7 (11)  19 (23)  12 (19)  80 (72)  16 (18)  Osteoporosis  27 (30)  2 (3)  24 (29)  9 (14)  18 (16)  25 (28)  Osteoarthritis  27 (30)  30 (45)  47 (57)  15 (24)  34 (31)  46 (52)  RA  0 (0)  3 (5)  34 (41)  0 (0)  12 (11)  2 (2)  Dementia  10 (11)  8 (12)  16 (19)  1 (2)  7 (6)  5 (6)  Depression  4 (4)  10 (15)  61 (73)  10 (16)  1 (1)  37 (42)  Thyroid disease  4 (4)  3 (5)  4 (5)  6 (10)  7 (6)  13 (15)  Non-skin cancer  0 (0)  5 (8)  6 (7)  59 (94)  4 (4)  16 (18)  Melanoma  4 (4)  3 (5)  5 (6)  11 (17)  6 (5)  2 (2)    Cross-sectional associations with cluster membership Mean (SE); Beta (95%CI) for adjusted model  Prescribed medications, n  3.1 (0.4); Reference  5.5 (0.4); 2.50 (1.37–3.63)**  5.6 (0.4); 2.58 (1.48–3.68)**  4.4 (0.4); 1.31 (0.17–2.45)*  5.6 (0.3); 2.53 (1.55–3.51)**  6.5 (0.4); 3.48 (2.44–4.53)**    n (%); OR (95%CI) for adjusted model  PIMS  15 (20%); Reference  13 (25%); 1.26 (0.66–2.43)  24 (39%); 1.80 (1.04–3.10)*  11 (21%); 1.03 (0.52–2.03)  29 (32%); 1.55 (0.90–2.65)  21 (31%); 1.56 (0.88–2.76)  PPOS  23 (30%); Reference  25 (48%); 1.58 (1.02–2.45)*  38 (62%); 2.00 (1.34–2.97)**  22 (42%); 1.38 (0.86–2.21)  49 (53%); 1.76 (1.19–2.59)**  40 (59%); 1.85 (1.24–2.74)**  Pre 12-month admission  15 (17%); Reference  24 (36%); 1.98 (1.13–3.47)*  25 (30%); 1.68 (0.95–2.96)  18 (29%); 1.59 (0.87–2.91)*  36 (33%); 1.82 (1.06–3.11)*  44 (49%); 2.72 (1.63–4.53)**    Longitudinal outcomes of cluster membership n (%); HR (95%CI) for adjusted model  48-month any hospitalisationa  57 (65%); Reference  55 (83%); 1.32 (0.90–1.94)  71 (86%) 1.48 (1.03–2.12)*  48 (76%); 1.11 (0.75–1.65)  90 (82%); 1.30 (0.92–1.83)  70 (79%); 1.34 (0.94–1.92)  48-month mortalityb  11 (12%); Reference  17 (26%); 2.15 (0.98–4.74)  26 (31%); 2.66 (1.26–5.59)*  18 (29%); 2.32 (1.06–5.09)*  33 (30%); 2.55 (1.24–5.25)*  28 (32%); 2.77 (1.34–5.73)**    Longitudinal outcomes of number of conditions    <3 conditions, 231/501  ≥3 conditions, 270/501  48-month any hospitalisationa  168 (73%): Reference  223 (83%); 1.10 (0.88–1.36)  48-month mortalityb  38 (16%): Reference  95 (35%); 2.24 (1.52–3.29)**  CAD, coronary artery disease; CHF, congestive heart failure; TIA, transient ischaemic attack; PVD, peripheral vascular disease; DM, Type II diabetes mellitus; AF, atrial fibrillation; COPD, chronic obstructive pulmonary disease; RA, rheumatoid arthritis; SE, standard error; CI, confidence interval. The bold values highlight prevalences of conditions, not stats significance values. Missing data: Number for prescribed medications, PIMs, PPOs: 100/501 (20%); hospital admission: 3/501 (1%). Note: The prevalence of hypertension in Clusters 1 to 6: 85%, 77%, 81%, 83%, 83%, 93%, 84%. Base model to examine longitudinal outcomes includes gender, education, NZ deprivation index. aBase model plus pre-hospital admission, weighted survival time. bBase model plus 48-month admission. *P < 0.05; **P < 0.01. In assessing the stability of the cluster solution, dendrograms and scree plots from the random sample showed similar patterns and prevalence of conditions in clusters indicating the overall cluster solution had a high degree of stability. We identified six distinct clusters for Māori and non-Māori (Tables 1 and 2) and labelled as described in the method section, e.g. for Māori, Cluster 3 was labelled ‘arthritis’ as 84% of members had arthritis, and Cluster 4 was labelled ‘CVD, respiratory and mental health’ as CVD and asthma/chronic obstructive pulmonary disease (COPD) were highly prevalent and this cluster had the highest prevalence of depression across the clusters (38%). Note is made that the prevalence of some conditions in the ‘Well’ cluster is higher than might be expected from its given label (relative to prevalence in other clusters). Patterns of conditions and outcomes in Māori There were more women than men in all clusters except for the ‘CHF-AF’ cluster. Compared to the ‘Well’ cluster, members in other clusters had significantly more medications and higher odds of having a PPO (Table 1). Participants the ‘arthritis’ and ‘diabetes’ clusters had higher risk of hospitalisation than ‘Well’ cluster members. Overall, participants in ‘diabetes’ and ‘complex multi-morbidity’ clusters had the worst outcomes, 86% and 91% having at least one hospitalisation and 41% and 39% 48-month mortality. In contrast to the earlier analyses of <3 versus ≥3 conditions, these clusters showed greater differentiation of the risk of hospitalisation and mortality. Patterns of conditions and outcomes in non-Māori The clusters in non-Māori and Māori differed but were labelled as closely as possible to facilitate interpretation (Table 2). Compared to cluster ‘Well’, other clusters had significantly more medications and a higher risk of a PPO (except for cluster ‘cancer’). All clusters, except ‘CHF-AF’, had more than twice the risk of all-cause mortality compared with the ‘Well’ cluster members. Overall, those in cluster ‘depression and arthritis’ had the worst outcomes: 86% having at least one hospitalisation and 31% 48-month mortality. Diabetes did not cluster in one group and was more evenly spread through Clusters 2–6. Depression clustered with arthritis amongst non-Māori, and participants with stroke were mainly in Cluster 6 (‘Stroke’). Combinations of conditions, rather the count (<3 versus ≥3), showed greater differentiation of the risk of hospitalisation and mortality. Discussion This is the first study to our knowledge to follow hospitalisation and mortality associated with patterns of multi-morbidity in octogenarians. This analysis has highlighted that profiles of conditions carry stronger associations with cross-sectional and longitudinal outcomes than the sum of those conditions. In considering the six patterns of conditions in Māori and non-Māori, several clinical insights are evident. CVD rarely occurred in isolation; 96% of those with CVD had co-morbidities. Conditions that consistently clustered together in both cohorts were ‘CHF and AF’; in non-Māori, with osteoarthritis; in Māori, with depression and COPD. Outcomes for participants with atrial fibrillation (AF) differed whether they were in the ‘CHF-AF’ cluster (same mortality as the ‘Well’ group), or in the ‘Complex multi-morbidity’ cluster (2.58 times higher than the ‘Well’ group). Combinations and outcomes are ethnic specific. Attempts to understand multi-morbidity have examined varying populations [8, 19] but lack of standard definitions and processes hamper consistent conclusions [10]. As well as differentiating likelihood of poor outcomes, this analysis has highlighted the frequent combination of depression and respiratory disease for 25% of Māori octogenarians. These two conditions could be linked through lower activity from COPD contributing to depression, or lower COPD medication adherence because of depression [20]. Depression is associated with higher readmissions rates for COPD [21] and lower health-related QoL [22]. In contrast, depression clustered with arthritis in non-Māori and this cluster had particularly low quality prescribing and high mortality. We add to the NZ experience and international literature on mental and physical comorbidity showing that morbidity of mental and physical health predicted increased risk of hospitalisation and mortality. While there are calls to manage depression and other concurrent medical conditions [21] specific intervention development based on patterns of conditions and targeted trials may be needed to establish how best to improve outcomes. Mechanisms to mortality may be mediated through medication appropriateness which varied between clusters for both Māori and non-Māori. For both ethnic groups, PPOs were more common than PIMs and high mortality appeared to track more closely with high PPOs. PIMs rates have been associated with adverse drug events [23], but PPOs to date have not been linked to poorer outcomes. Prescribing patterns in Māori and non-Māori are explored further in another manuscript (under review). Of note, the criteria for quality of prescribing was based on current single disease guidelines. Further work is needed in understanding whether PIMs or PPOs are more relevant in predicting outcomes. The idea of ‘personalised medicine’ has gained credence internationally in recent years, frequently in the context of using molecular analysis to customisation healthcare in specific populations. The ageing population is heterogonous, and examining multi-morbidity patterns may prompt better ‘personalised care’. Smith et al. suggested that an integrated care management process comprising organisational (e.g. care co-ordination/management, regulatory interventions), pharmacologic and non-pharmacologic interventions (e.g. education, exercise therapy, cognitive behavioural therapy) focusing on specific area(s) is needed [24]. Further calls for precision in assessment of outcomes challenges current conceptualisations of management [25]. We add that a greater understanding of common patterns of conditions and their associated outcomes is needed. To complement ‘personalised care’, the healthcare system and clinical practice also needs to be cognisant of and embrace bi-cultural (e.g. NZ) and multi-cultural (e.g. USA, UK) older populations supporting healthcare professionals to be culturally competent in delivering care. We question whether multi-morbidity patterns may be consistent in differing populations. Very few studies have examined octogenarians, and those that have, have included hypertension as present in several clusters due to its high prevalence [8]. We excluded hypertension and thus potentially allowed underlying groupings to be identified. Some clusters identified here are similar to others in that coronary artery disease (CAD) and congestive heart failure (CHF) cluster together for Māori (‘complex multi-morbidity’), also observed in those aged 75+ in Sweden [26], and in the oldest old in Sweden dementia clustered with depression [26] as it did for both our cohorts. We did not identify cardio-metabolic clusters, common in other studies [8]. In this study when diabetes predominated, CVD was not co-prevalent in either ethnic group. Differences may be explained by different diagnostic groups included in studies and without a consistent platform populations cannot be compared. Larger studies with defined diagnostic groupings need to replicate this result. Further work should attempt consistency in multi-morbidity studies and develop specific treatment combinations related to common patterns. Other studies have used different techniques to identify clustering of conditions [8] (e.g. multiple correspondence analysis [27]; factor analysis [28, 29]; hierarchical cluster analysis by clustering of conditions [18]). Our clusters differed somewhat as we took a patient-centred approach and identified clusters of participants (rather than conditions) which could potentially lead to development of specific treatment regimes for specific groupings of co-existing conditions. In addition to different analytical techniques, different populations (age, ethnicity, geographical locations) and methods in ascertaining diagnosis (number and type of conditions) define other results [8, 27–30]. While the experience of NZ’s indigenous people is likely to be similar to those of indigenous peoples in other resource-rich countries with similar colonial histories, other indigenous peoples may have different histories and contemporary circumstances. Other ethnic diaspora ageing in a similar or new cultures/value system may also benefit from a separate examination of multi-morbidity. Strengths and limitations The strengths of this analysis are the certainty with which medications and diagnoses were identified, the inclusion of a representative group of the older indigenous population and the robustness of revalidated analysis. Limitations include the less stringent criteria (versus factor analysis) in deciding number of clusters in hierarchical clustering analysis [18]. Subjective criteria based on subject expertise were frequently employed in deciding number of clusters to ensure results are interpretable and meaningful, i.e. specificity to ensure manageability versus generalisability, and to optimise strategic attention to groupings of co-existing conditions. We pre-defined clinical criteria in deciding number of clusters. The low response rate (57%) [10] could be a potential source of bias (e.g. the frailer being less likely to participate). We included ‘ever’ diagnoses rather than ‘current’ diagnoses and this mainly affected the cancer group and perhaps the depression diagnosis. Missing data on outcome variables may lead to under estimation of associations between outcomes and clusters, a Type II error. This is less likely for hospitalisations and mortality with a National Health Index number, but interview data for medications was not universally available. These findings will better inform the implementation of health programming and policies for healthcare delivery for the aged through emphasis on specific patterns rather than individual diseases in educational programmes, stimulation of specific trial research on outcomes related to different treatment regimens; and through raising public awareness of the complexity of multi-morbidity. Conclusions We have shown that certain disease combinations put older people at particular risk of poor outcome. These findings support personalised care for different ethnic groups based on differences between the patterns of multi-morbidity and observed associated outcomes. Further work will examine treatment patterns and whether and how they associate with outcomes within these combinations of multi-morbidity. Key points Clusters of conditions provide better differentiation of hospitalisation and mortality risk than count of conditions. Different ethnic groups have distinctive shared conditions with Cardiovascular Disease (CVD). Better ‘personalised care’ requires examining patterns of multi-morbidity. Supplementary data Supplementary data mentioned in the text are available to subscribers in Age and Ageing online. Funding This work was supported by the Health Research Council of New Zealand (HRC 09/068B; UoA ref: 3624940) and Ministry of Health New Zealand (MOH ref: 345426/00; UoA ref 3703221) which funded the project management and data collection work; Ngā Pae o te Māramatanga (UoA ref: 3624946) which funded the Māori engagement and project management; New Zealand Heart Foundation project grant for investigating cardiac markers (UoA Ref: 3625921) and a Heart Foundation Research Fellowship (UoA ref: 3702288). We thank the sponsors. Conflict of interest None. Acknowledgements We wish to acknowledge the participants, their families and whanau for supporting the study. We are extremely grateful to Dr. Lorna Dyall (PhD) and Dr. Mere Kepa (EdD) for the significant contributions in drafting this manuscript who are now enjoying their retirement days. Their inputs on interpretation of data from Māori worldview are instrumental. We thank Te RōpuKaitiaki o ngā tikanga Māori for their guidance. We acknowledge the community partners in LiLACS NZ who engaged with participants and collected data (Western Bay of Plenty Primary Health Organisation, Ngā Matāpuna Oranga Kaupapa Māori Primary Health Organisation, Te Korowai Aroha Trust, Te Rūnanga o Ngati Pikiao, Rotorua Area Primary Health Services, Ngati Awa Research & Archives Trust, Te Rūnanga o Ngati Irapuaia and Te Whanau a Apanui Community Health Centre). References 1 Barnett K, Mercer SW, Norbury M et al.  . Epidemiology of multimorbidity and implications for health care, research, and medical education: a cross-sectional study. Lancet  2012; 380: 37– 43. Google Scholar CrossRef Search ADS PubMed  2 Fortin M, Hudon C, Haggerty J et al.  . Prevalence estimates of multimorbidity: a comparative study of two sources. BMC Health Serv Res  2010; 10: 111. Google Scholar CrossRef Search ADS PubMed  3 Aarts S, den Akker M.v., Bosma H et al.  . The effect of multimorbidity on health related functioning: Temporary or persistent? Results from a longitudinal cohort study. J Psychosom Res  2012; 73: 211– 7. Google Scholar CrossRef Search ADS PubMed  4 Aarts S, van den Akker M, Hajema KJ et al.  . Multimorbidity and its relation to subjective memory complaints in a large general population of older adults. Int Psychogeriatr  2011; 23: 616– 24. Google Scholar CrossRef Search ADS PubMed  5 Fortin M, Dubois M-F, Hudon C et al.  . Multimorbidity and quality of life: a closer look. Health Qual Life Outcomes  2007; 5: 52. Google Scholar CrossRef Search ADS PubMed  6 van den Bussche H, Schon G, Kolonko T et al.  . Patterns of ambulatory medical care utilization in elderly patients with special reference to chronic diseases and multimorbidity—results from a claims data based observational study in Germany. BMC Geriatr  2011; 11: 54. Google Scholar CrossRef Search ADS PubMed  7 Fortin M, Stewart M, Poitras M-E et al.  . A systematic review of prevalence studies on multimorbidity: toward a more uniform methodology. Ann Fam Med  2012; 10: 142– 51. Google Scholar CrossRef Search ADS PubMed  8 Violan C, Foguet-Boreu Q, Flores-Mateo G et al.  . Prevalence, determinants and patterns of multimorbidity in Primary Care: a systematic review of observational studies. PLOS one  2014; 9: e102149. Google Scholar CrossRef Search ADS PubMed  9 Kerse N, Teh R, Moyes S et al.  . Cohort profile: Te Puawaitanga o Nga Tapuwae Kia Ora Tonu, Life and Living in Advanced Age: a Cohort Study in New Zealand (LiLACS NZ). Int J Epidemiol  2015; 44: 1823– 32. Google Scholar CrossRef Search ADS PubMed  10 Dyall L, Kepa M, Hayman K et al.  . Engagement and recruitment of Māori and non-Māori people of advanced age to LiLACS NZ. Aust N Z J Public Health  2013; 37: 124– 31. Google Scholar CrossRef Search ADS PubMed  11 Teh R, Doughty R, Connolly M et al.  . Agreement between self-reports and medical records of cardiovascular disease in octogenarians. J Clin Epidemiol  2013; 66: 1135– 43. Google Scholar CrossRef Search ADS PubMed  12 World Health Organisation. World Health Organisation Collaborating Centre for Drug Statistics Methodology. [cited 2011; Available from: http://www.whocc.no/. 13 Gallagher P, Ryan C, Byrne S et al.  . STOPP (Screening Tool of Older Person’s Prescriptions) and START (Screening Tool to Alert doctors to Right Treatment). Consensus validation. Int J Clin Pharmacol Ther  2008; 46: 72– 83. Google Scholar CrossRef Search ADS PubMed  14 Teh R, Kerse N, Kepa M et al.  . Self-rated health, health related behaviours and medical conditions of Māori and non-Māori in advanced age: LiLACS NZ. N Z Med J  2014; 127: 13– 29. Google Scholar PubMed  15 Hill S, Sarfati D, Blakely T et al.  . Survival disparities in Indigenous and non-Indigenous New Zealanders with colon cancer: the role of patient comorbidity, treatment and health service factors. J Epidemiol Community Health  2010; 64: 117– 23. Google Scholar CrossRef Search ADS PubMed  16 Chan WC, Wright C, Riddell T et al.  . Ethnic and socioeconomic disparities in the prevalence of cardiovascular disease in New Zealand. NZMJ  2008; 121: 11– 20. 17 Pihama L. A conversation about kaupapa Māori theory and research. In: Hutchings J, Potter H, Taupo K, eds. New Zealand Council for Educational Research: Challenges of Kaupapa Māori Research . Wellington: Pipitea marae, 2011; pp. 49– 55. 18 Cornell JE, Pugh JA, Williams JW Jr et al.  . Multimorbidity clusters: clustering binary data from multimorbidity clusters: clustering binary data from a large administrative medical database. Appl Multivar Res  2009; 12: 163– 82. Google Scholar CrossRef Search ADS   19 Sinnige J, Korevaar JC, Westert GP et al.  . Multimorbidity patterns in a primary care population aged 55 years and over. Fam Pract  2015; 32: 505– 13. Google Scholar CrossRef Search ADS PubMed  20 Bender BG. Nonadherence in chronic obstructive pulmonary disease patients: what do we know and what should we do next? Curr Opin Pulm Med  2014; 20: 132– 7. Google Scholar CrossRef Search ADS PubMed  21 Pooler A, Beech R. Examining the relationship between anxiety and depression and exacerbations of COPD which result in hospital admission: a systematic review. Int J Copd  2014; 9: 315– 30. Google Scholar CrossRef Search ADS   22 Sundh J, Johansson G, Larsson K et al.  . Comorbidity and health-related quality of life in patients with severe chronic obstructive pulmonary disease attending Swedish secondary care units. Int J Copd  2015; 10: 173– 83. Google Scholar CrossRef Search ADS   23 Hamilton H, Gallagher P, Ryan C et al.  . Potentially inappropriate medications defined by STOPP criteria and the risk of adverse drug events in older hospitalized patients. Arch Int Med  2011; 171: 1013– 9. Google Scholar CrossRef Search ADS   24 Smith SM, Soubhi H, Fortin M et al.  . Managing patients with multimorbidity: systematic review of interventions in primary care and community settings. Br Med J  2012; 345: e5205. Google Scholar CrossRef Search ADS   25 Bierman AS, Tinetti ME. Precision medicine to precision care: managing multimorbidity. Lancet  2016; 388: 2721– 3. Google Scholar CrossRef Search ADS PubMed  26 Marengoni A, Rizzuto D, Wang HX et al.  . Patterns of chronic multimorbidity in the elderly population. J Am Geriatr Soc  2009; 57: 225– 30. Google Scholar CrossRef Search ADS PubMed  27 Garcia-Olmos L, Salvador CH, Alberquilla A et al.  . Comorbidity patterns in patients with chronic diseases in general practice. PLoS ONE  2012; 7: e32141. Google Scholar CrossRef Search ADS PubMed  28 Schäfer I, von Leitner E-C, Schön G et al.  . Multimorbidity patterns in the elderly: a new approach of disease clustering identifies complex interrelations between chronic conditions. PLoS ONE  2010; 5: e15941. Google Scholar CrossRef Search ADS PubMed  29 Kirchberger I, Meisinger C, Heier M et al.  . Patterns of multimorbidity in the aged population. Results from the KORA-Age study. PLoS ONE  2012; 7: e30556. Google Scholar CrossRef Search ADS PubMed  30 Steinman MA, Lee SJ, Boscardin WJ et al.  . Patterns of multimorbidity in elderly veterans. J Am Geriatr Soc  2012; 60: 1872– 80. Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2017. Published by Oxford University Press on behalf of the British Geriatrics Society.All rights reserved. For permissions, please email: journals.permissions@oup.com

Journal

Age and AgeingOxford University Press

Published: Mar 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 12 million articles from more than
10,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Unlimited reading

Read as many articles as you need. Full articles with original layout, charts and figures. Read online, from anywhere.

Stay up to date

Keep up with your field with Personalized Recommendations and Follow Journals to get automatic updates.

Organize your research

It’s easy to organize your research with our built-in tools.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

Monthly Plan

  • Read unlimited articles
  • Personalized recommendations
  • No expiration
  • Print 20 pages per month
  • 20% off on PDF purchases
  • Organize your research
  • Get updates on your journals and topic searches

$49/month

Start Free Trial

14-day Free Trial

Best Deal — 39% off

Annual Plan

  • All the features of the Professional Plan, but for 39% off!
  • Billed annually
  • No expiration
  • For the normal price of 10 articles elsewhere, you get one full year of unlimited access to articles.

$588

$360/year

billed annually
Start Free Trial

14-day Free Trial