Background: Residence in public housing, a subsidized and managed government program, may affect health and healthcare utilization. We compared healthcare use in the year before individuals moved into public housing with usage during their first year of tenancy. We also described trends in use. Methods: We used linked population-based administrative data housed in the Population Research Data Repository at the Manitoba Centre for Health Policy. The cohort consisted of individuals who moved into public housing in 2009 and 2010. We counted the number of hospitalizations, general practitioner (GP) visits, specialist visits, emergency department visits, and prescriptions drugs dispensed in the twelve 30-day intervals (i.e., months) immediately preceding and following the public housing move-in date. Generalized linear models with generalized estimating equations tested for a period (pre/post-move-in) by month interaction. Odds ratios (ORs), incident rate ratios (IRRs), and means are reported along with 95% confidence intervals (95% CIs). Results: The cohort included 1942 individuals; the majority were female (73.4%) who lived in low income areas and received government assistance (68.1%). On average, the cohort had more than four health conditions. Over the 24 30-day intervals, the percentage of the cohort that visited a GP, specialist, and an emergency department ranged between 37.0% and 43.0%, 10.0% and 14.0%, and 6.0% and 10.0%, respectively, while the percentage of the cohort hospitalized ranged from 1.0% to 5.0%. Generally, these percentages were highest in the few months before the move-in date and lowest in the few months after the move-in date. The period by month interaction was statistically significant for hospitalizations, GP visits, and prescription drug use. The average change in the odds, rate, or mean was smaller in the post-move-in period than in the pre-move-in period. Conclusions: Use of some healthcare services declined after people moved into public housing; however, the decrease was only observed in the first few months and utilization rebounded. Knowledge of healthcare trends before individuals move in are informative for ensuring the appropriate supports are available to new public housing residents. Further study is needed to determine if decreased healthcare utilization following a move is attributable to decreased access. Keywords: Public housing, Healthcare use, Health services, Health status, Record linkage, Administrative data * Correspondence: email@example.com Department of Community Health Sciences, University of Manitoba, S113-750 Bannatyne Ave, Winnipeg, Manitoba R3E 0W3, Canada Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Hinds et al. BMC Health Services Research (2018) 18:411 Page 2 of 13 Background which may improve health, at least temporarily. We sus- Public housing is a form of subsidized housing owned pected that healthcare use may not remain at a reduced and/or managed by government (municipal, provincial/ level because public housing residents often have state, or federal). Public housing tenants pay rent geared chronic conditions which require ongoing care. to income - - usually 30% of the household income . The intent of public housing is to offer a broad safety Methods net to the economically disadvantaged . Study cohort Public housing residents tend to be in poorer health Manitoba is an ethnically diverse Canadian province compared to the general population, with lower self- with a population of approximately 1.3 million, with 55% reported health [3–5], a higher prevalence of chronic dis- residing in the City of Winnipeg, the capital. The cohort eases (including diabetes, hypertension, asthma) [3, 6–8], included all adults (18+ years) who moved into public injuries , and mental health disorders [8–11]. Public housing provided by the provincial ministry, Manitoba housing residents are more likely to engage in risky health Housing, between January 1, 2009 and December 31, behaviors, including smoking, alcohol and drug use, as 2010 and were listed as the primary applicant. There are well as risky sexual behaviors [3, 4, 12–17], and generally approximately 35,000 social housing units in Manitoba; have lower levels of physical activity [5, 18–24]. There is of which approximately 13,000 are public housing units evidence, however, that residents’ poor health precedes spread throughout the province that are directly man- their application for public housing . aged by Manitoba Housing . More than 30,000 indi- Research about healthcare use among public housing viduals reside in public housing units in a year, with residents is limited and inconsistent. McNeill et al. (2009) approximately half under the age of 20 years . found that 87% of their sample of 1554 individuals resid- The cohort included individuals registered with the ing in 12 public housing sites in Massachusetts reported Manitoba Health Services Insurance Plan in the year having access to a regular health care provider . In prior to and in the year following their public housing Black et al.’s two studies [27, 28], 90% of the older adult move-in date. Only new applicants were included; indi- participants had received medical care in the past six viduals residing in public housing within two years of months, averaging six medical visits during this time, but their 2009/2010 move-in date were excluded. Manitoba only 43% of residents had obtained care from a private Housing requires that people reapply if they wish to physician (the rest obtained care from a clinic or hospital- switch to a different housing unit; therefore, these based provider). After adjusting for demographic, health, people do not have a year out of public housing before hospital, and neighbourhood characteristics, children in their 2009/2010 move-in date. In Churchill, a remote public housing that had not been redeveloped were signifi- northern Manitoba community, public housing is used cantly more likely to have recurrent use of acute care ser- to supplement the shortage of affordable market hous- vices when compared with children not residing in public ing. Residents of that community were excluded as it is housing or in public housing that had been renovated and not possible to distinguish between those paying market redeveloped as part of the Housing Opportunities for rate rents and those living in subsidized units . Indi- People Everywhere (HOPE VI) program . viduals living in public housing for less than one year We endeavored to determine if healthcare use changes were also excluded. over the short-term among new residents of public housing. Our research objective was to examine trends Data source in healthcare use one year before and one year after in- The Population Research Data Repository housed at the dividuals move into public housing. We tested for a Manitoba Centre for Health Policy is a rich collection of change in healthcare use between the two periods. We anonymized health and social administrative databases hypothesized that all forms of healthcare use would in- linkable at the individual level via a unique scrambled crease before the move-in date since health may be re- personal health identification number. Previous re- lated to applying. In a previous study, we found that searchers have identified all residents who applied to people who applied to public housing were more likely and/or who moved into public housing and linked this to have health conditions and were higher users of information to a comprehensive set of health and socio- healthcare services than individuals who were similar in economic indicators and outcomes [8, 25, 31]. terms of their socioeconomic characteristics . We The Tenant Management System (TMS) was used to also hypothesized that all forms of healthcare use would indicate residents of Manitoba Housing’s rental housing. decrease after the move-in date. The reasons for this de- The TMS contains information on public housing man- crease are potentially multifaceted and could include re- aged by the provincial government (approximately 2300 duced access to healthcare, an adjustment period after buildings and 13,000 units). The number of public housing moving in, or to a better housing and financial situation, units has remained fairly constant over time; approximately Hinds et al. BMC Health Services Research (2018) 18:411 Page 3 of 13 59% of the units are located in the Winnipeg health region and economic variables were defined for the 365 days . Demographic (e.g., sex, birth date, six-digit postal prior to the move-in date. Six-digit postal code was code) and health coverage information (e.g., start and end used to determine residence before a move to public of coverage dates) was obtained from the Population Regis- housing. Region of residence was defined as urban or try. The Registry contains this information for all Manitoba rural (i.e., Winnipeg or non-Winnipeg). Residential residents registered with the provincial health plan (ex- mobility was defined by identifying changes in postal cludes military personnel, the Royal Canadian Mounted code in the 365 days prior to the move-in date. In Police, and new residents) and is updated every six months Winnipeg, a postal code covers a medium sized apart- (June and December). The Social Assistance Management ment or a residential block, while postal code areas Information Network provides information on households are larger outside of the city. Individuals were classi- receiving financial support under the provincial Employ- fied as movers or non-movers depending on whether ment and Income Assistance program. Average household their postal code changed [36, 37]. The Manitoba income information from the 2006 Canadian Census was Housing move-in and move-out dates were used to used to create an area-level measure of income (i.e., in- determine length of tenancy. Application reason indi- come quintile). cated why individuals applied to Manitoba Housing Information on discharges from all acute and chronic and the move-out reasons were grouped into volun- care facilities was obtained from the hospital discharge ab- tary moves and eviction. stracts database. Up to 25 diagnosis codes are recorded The economic variables were receipt of income assist- using the International Classification of Diseases, 10th re- ance (IA) and income quintile (IQ). Individuals were vision, Canadian version (i.e., ICD-10-CA). The physician classified as recipients of IA if they or a member of their billing claims database captures all fee-for-service physi- household received any form of IA in the 365 days prior cian visits, which comprises the vast majority of visits. A to the move-in date . IA is based on financial need study from 2004 estimated that 93% of physicians in as well as other eligibility criteria. IQ is an area-level Winnipeg are remunerated on a fee-for-service basis , measure of income based on the average household in- while two recent studies estimated that 84.1% and 86.4% come in the dissemination area (DA), the smallest geo- of patients diagnosed with diabetes in Manitoba were graphic level for which Census data are reported . treated by a fee-for-service physician [33, 34]. The diagno- The DAs are sorted from poorest to wealthiest and sis deemed most responsible for the physician visit is re- grouped into quintiles. Each quintile represents approxi- corded using a three-digit International Classification of mately 20% of the population. Different household in- Diseases, 9th revision, Clinical Modification (i.e., ICD-9- come cut-offs define quintiles for urban and rural areas. CM) code. Information about visits to adult emergency Health status in the 365 days prior to the move-in date departments (EDs) in Winnipeg was obtained from the was established using diagnosis codes for selected condi- Emergency Department Information System (EDIS). The tions in the physician billing claims and hospital dis- EDIS database contains information on urgency of need charge abstracts (Appendix). These conditions have been for treatment, arrival and discharge status, timing of used previously to describe the health of public housing healthcare events in the ED (e.g., registration, triage, initi- applicants and/or residents [8, 25]. Mental health condi- ation of treatment), chief patient complaints, and diagnos- tions included schizophrenia, affective (mood and anx- tic and blood tests . ICD codes are not recorded in the iety) disorders, and substance abuse disorders. Physical EDIS database and there is no corresponding data avail- health conditions included respiratory illness (e.g., able on ED visits outside of Winnipeg. Information about asthma, chronic obstructive pulmonary disease, bron- prescription drugs dispensed from community pharma- chitis, emphysema), diabetes, hypertension, cancer, arth- cies, including the dispensation date, drug identification ritis, and injuries. Health status in the 365 days prior to number, and dosage, was obtained from the Drug Pro- the move-in date was summarized using Aggregated gram Information Network (DPIN) database. The DPIN Diagnostic Groups (ADGs) [40, 41]. ADGs are groups of database includes some over-the-counter medications. A ICD-9-CM/ICD-10-CA codes that represent diagnoses more detailed description of all of the databases is avail- that are clinically similar and for which the expected or able on the Manitoba Centre for Health Policy website actual use of health care services is similar. The John (http://umanitoba.ca/faculties/health_sciences/medicine/ Hopkins Adjusted Clinical Group® (ACG®) Case-Mix units/chs/departmental_units/mchp/). System version 9 clusters the ICD codes into 32 mutu- ally exclusive ADGs. A higher ADG score indicates a Study variables greater level of comorbidity. Demographic variables included sex and age group To examine trends in healthcare use, the number of (18–24, 25–39, 40–64, 65+ years) and were defined at general practitioner (GP) physician visits and the the move-in date. Geographic, residential mobility, number of specialist visits were summarized for Hinds et al. BMC Health Services Research (2018) 18:411 Page 4 of 13 twelve 30-day intervals (i.e., months) before and after the visit, hospitalization/visit in the 30-day interval). The re- move-in date; these were determined using the date of ser- gression coefficients are presented as odds ratios (ORs) vice and physician type in the physician billings claims along with their corresponding 95% confidence intervals database. The number of inpatient hospitalizations in each (95% CIs). A negative binomial distribution, appropriate 30-day interval was calculated using the admission and for counts of relatively rare events that exhibit extra- discharge dates recorded in the hospital discharge ab- Poisson variation, was adopted for modeling the number stracts database. Pregnancy-related hospitalizations were of GP visits. The regression coefficients are presented as not included. Hospitalizations within 24 hours and incident rate ratios (IRRs) along with their corresponding transfers between facilities were considered a single 95% CIs. The number of prescription drugs was modeled hospitalization . Two hospitalization measures were using a normal distribution and the regression coeffi- defined. First, hospital stays that spanned more than one cients are presented as means. All analyses were con- 30-day interval were counted in each interval to account ducted using SAS software version 9.4 (SAS Institute for variations in length of stay. For the second measure, a Inc., Cary NC, USA). hospital stay was counted only in the interval in which an individual was admitted to hospital. The number of differ- Results ent prescription drugs using the fourth-level of the Ana- After applying the exclusions, the cohort comprised tomical Therapeutic Chemical (ATC) classification system 1942 (46.4%) of the 4183 adult primary applicants to in each 30-day interval was determined from the dispensa- Manitoba’s rental housing who moved in between Janu- tion date in the DPIN database. The fourth-level ATC code ary 1, 2009 and December 21, 2010 (Fig. 1). The largest denotes the chemical, therapeutic, or pharmacological exclusion (28.9%) was because individuals had resided in subgroup and has been used by other researchers to count public housing in the two years before their 2009/2010 the number of different drugs . The number of days on move-in date, the majority of whom were continuous each drug was determined from the days’ supply and was residents (i.e., reapplied to move to a different unit) or used to determine if use spanned more than one interval. had moved out and then moved back after a short Prescription drugs spanning more than one interval were period of time (i.e., less than one year). There were less counted in each interval. The number of ED visits in each than six individuals with multiple housing records who 30-day interval was also calculated. ED visits overlapping had more than a one year break from residing in public more than one interval were counted in each interval. housing before their 2009/2010 move-in date. Statistical analysis Cohort sociodemographic, health, and housing Descriptive statistics, including means, standard devia- characteristics tions, and frequencies were used to characterize the co- The sociodemographic and health characteristics of the hort. We used regression models to test for linear trends cohort are reported in Table 1. Almost three-quarters of in the percentages/means of the healthcare measures over the cohort were female and the average age at the move- time. Generalized linear models with generalized estimat- in date was 41.7 years (SD = 18.6). The majority were ing equations (GEEs) tested for changes in healthcare. We urban residents (54.5%). An IQ gradient showed that as adopted an unstructured correlation structure, the least neighbourhood income increased, the percentage of the restrictive structure, to account for the within-subject cor- cohort residing in those areas decreased. More than relation over the 30-day intervals. The quasi-likelihood in- two-thirds (68.2%) of households received IA. Approxi- formation criterion (QIC) was used to assess model fit mately one-third of the cohort reported an address . Unadjusted and adjusted models were fit to the data. change in the year prior to the move-in date. The most The unadjusted models included period (pre- and post- common physical and mental health conditions were arth- move-in date), month (30-day interval), and the period by ritis (26.3%) and affective disorders (32.9%), respectively. month two-way interaction. The adjusted models included The average number of ADGs was 4.52 (SD = 3.14). the period, month, and the period by month two-way The housing-related characteristics of the cohort are interaction as well as demographic (i.e., sex, age group), reported in Table 2. Health/medical reasons were the geographic (i.e., region of residence), economic (i.e., IQ, fourth most common reasons for applying to public receipt of IA), residential mobility, and health status char- housing. During the observation period (January 1, 2009 acteristics (i.e., physician-diagnosed mental (schizophre- to March 31, 2013), one-third of the cohort moved out nia, mood disorders, substance abuse disorders) and of public housing. Among the movers, 21.9% were physical health (i.e., injury, diabetes, respiratory illness, evicted and 78.1% moved out voluntarily. The median arthritis, cancer, hypertension) conditions and ADGs). length of time between the application date and the ap- Hospitalizations, specialist visits, and ED visits were proval date was 18 days (Q1 = 8, Q3 = 45) and the me- modeled as dichotomous variables (i.e., no hospitalization/ dian length of time between the approval date and the Hinds et al. BMC Health Services Research (2018) 18:411 Page 5 of 13 Fig. 1 Flow chart for construction of the study cohort move-in date was 55 days (Q1 = 28, Q3 = 120). In total, ranged between 10.0% and 14.0%; the percentages were the median length of time between the application date lower after the move-in date. However, there was no evi- and the move-in date was 89 days (Q1 = 47, Q3 = 180). dence that the linear trend in the percentage of specialist visits differed in the two periods, F(1,20) = 0.3, p =0.56. Healthcare utilization On average, the cohort had 2.19 specialist physician visits Healthcare use in the twelve 30-day intervals before (SD = 4.11, median = 0) in the year before the move-in and after the public housing move-in date is shown in date and 1.92 specialist physician visits (SD = 3.87, Figs. 2 and 3. Additional file 1: Table S1 shows the per- median = 0) in the year after the move-in date. centages/means along with the 95% confidence inter- The percentages of the cohort hospitalized increased vals for the healthcare use measures in each of the 24 before the move-in date, peaking three months before 30-day intervals. As presented in Fig. 2, in any 30-day the move-in date at approximately 4.0%, and then de- interval, between 37.0% and 43.0% of the cohort visited creased after the move-in date and stabilized at around aGP. In themonth afterthe move-indatethe percent- 2.0% at six months after the move-in date. There was age of the cohort who visited a GP decreased, but then evidence that the linear trend in the percentages in the fluctuated at an increased level comparable to the pre- pre-move-in period differed significantly from the linear move-in percentages. There was no evidence that the trend in the post-move-in period, F(1, 20) = 31.4, p <0.01. linear trend in the percentage of GP visits differed in There was a similar pattern in the percentage hospitalized the two periods, F(1,20) = 2.8, p = 0.11). The cohort av- when hospitalizations were only counted in the period in eraged 7.42 GP visits (SD = 7.34, median = 6) in the year which a person was admitted, except in the two months before and 7.33 GP visits (SD = 6.95, median = 6) in the before the move-in date. The percentage admitted to hos- year after the move-in date. The percentage of the co- pital peaked three months before the move-in date (3.2%) hort visiting a specialist physician in a 30-day interval and then decreased, while the percentage hospitalized Hinds et al. BMC Health Services Research (2018) 18:411 Page 6 of 13 Table 1 Sociodemographic and health characteristics of the percentage of ED visits differed in the two periods, cohort in the 365 days prior to the public housing move-in date F(1,20) = 0.5, p = 0.50. In total, 46.7% and 43.4% of the (N = 1942) Winnipeg residents visited an emergency department Variables Categories N % in the year before and year after the move-in date, re- Sex Males 517 26.6 spectively. There is no data available on visits to an emergency department outside of Winnipeg. Females 1425 73.4 As shown in Fig. 3, the mean number of different pre- Age (years) at Move-in Date 18–24 385 19.8 scriptions filled in a 30-day interval increased over time, 25–39 678 34.9 from two prescriptions the year before the move-in date 40–64 621 32.0 to three prescriptions at the end of the first year in pub- 65+ 258 13.3 lic housing. As shown in Additional file 1: Table S1, the Region Winnipeg 1058 54.5 95% CIs for the means at the beginning of the pre- move-in period do not overlap with the 95% CIs for the Non-Winnipeg 884 45.5 means at the end of the post-move-in period and there Income Quintile Q1 (poorest) 832 42.8 was evidence that the linear trend in the mean number Q2 478 24.6 of different prescriptions filled differed in the two pe- Q3 319 16.4 riods, F(1,20) = 17.1, p < 0.01. On average, the cohort Q4 197 10.1 filled 6.38 (SD = 5.27, median = 5) different prescriptions Q5 (affluent) 93 4.8 in the year before the move-in date and 6.51 (SD = 5.11, median = 6) unique prescriptions in the year after the NF 23 1.2 move-in date. Income Assistance Yes 1323 68.1 The estimates and 95% confidence intervals (CIs) for No 619 31.9 the period by month interactions are presented in Change in Postal Code Yes 657 33.8 Table 3. The period by month interaction was statisti- No 1285 66.2 cally significant for hospitalizations (all intervals) (p <0. Physical Disorders Arthritis 459 26.3 01), GP visits (p = 0.05), and prescriptions (p < 0.01) in the adjusted models. The average change in the odds or Injury 425 21.9 rate of utilization was smaller in the post-move-in period Respiratory Disease 355 18.3 than in the pre-move-in period. Additional file 2: Table Hypertension 286 14.7 S2 shows the Chi-square test statistics and p-values for Diabetes 176 9.1 the main and interaction effects and Additional file 3: Ischemic Heart Disease 37 1.9 Table S3 shows the estimates and 95% CIs for the vari- Cancer 36 1.9 ables in each model. The absolute estimates and 95% CIs for each period Inflammatory Bowel Disease 11 0.6 are presented in Table 4. The odds of hospitalization in Mental Disorders Affective Disorders 638 32.9 the pre-move-in period changed over time (OR = 1.07; Substance Abuse Disorders 145 7.5 95% CI 1.04, 1.10), but there was no significant change Schizophrenia 61 3.1 in the odds in the post-move-in period. Similarly, the Note. NF Not Found odds of an ED visit increased significantly over time in the pre-move-in period (OR = 1.02; 95% CI 1.00, 1.04), (accounting for length of stay) remained at an ele- but the change in the odds was not statistically signifi- vated level in the three months before the move-in cant in the post-move-in period. In the pre-move-in date. There was no evidence that the linear trend dif- period, the GP visit rate increased slightly (IRR = 1.01, fered in the two periods, F(1, 20) = 2.9, p =0.10. In 95% CI 1.01, 1.02), but the change in the GP visit rate total, 16.9% of the cohort members were hospitalized was not statistically significant in the post-move-in in the year before the move-in date and 13.9% were period. For prescription medications, there was a statisti- hospitalized in the year after. cally significant change in both the pre-move-in period The percentage of Winnipeg residents who visited an (0.06; 95% CI 0.05, 0.07) and post-move-in period (0.02; emergency department fluctuated between 7.0% and 10. 95% CI 0.01, 0.03). 0% over the 24 30-day intervals. The percentage peaked in the three months prior to the move-in date, declined Discussion within the three months after the move-in date, but then This cohort of new public housing residents was primarily the percentages rebounded to the pre-move-in date comprised of female, urban residents who lived in very levels. There was no evidence that the linear trend in the low income areas, and received some form of government Hinds et al. BMC Health Services Research (2018) 18:411 Page 7 of 13 Table 2 Housing-related characteristics of the cohort (N = 1942) Variables Categories N % Application Reason Overcrowded conditions 632 32.5 Cannot afford current rent/utilities 516 26.6 Safety/security 209 10.8 Health/medical 140 7.2 Not specified 107 5.5 To be closer to family/ employment/education 98 5.1 Family separation 77 4.0 Physical condition is unsatisfactory 52 2.7 Notice to vacate 50 2.6 Unable to maintain current home/yard 33 1.7 To be closer to medical facilities 28 1.4 Moved-Out Status Moved out voluntarily 505 26.0 Evicted 142 7.3 Did not move out 1295 66.7 Time in Public Housing (days) Moved 734.1 (256.5) Moved out voluntarily 726.2 (255.0) Evicted 762.5 (260.5) Did not move out 1175.3 (210.9) Notes. Application reason was based on information provided prior to the move-in date. Move-out status and time in public housing were based on information obtained after the move-in date (between 2009 and 2013). Mean (Standard Deviation) Fig. 2 Percentage of the study cohort who were hospitalized, visited an emergency department (Winnipeg residents), and saw a GP or a specialist in the year before and the year after the public housing move-in date. Note. N = 1942 for GP visits, specialist visits, and hospitalizations. The N for ED visits (Winnipeg residents) fluctuated between 1067 and 1089 depending on who resided in Winnipeg in any given period Hinds et al. BMC Health Services Research (2018) 18:411 Page 8 of 13 Fig. 3 Mean number of prescriptions per 30-day interval filled by the study cohort (N = 1942) in the year before and the year after the public housing move-in date assistance. Approximately one-third changed their lo- in our cohort was high, which may be due to a high preva- cation of residence in the year prior to moving into lence of smoking as reported by other studies [4, 13, 15, public housing. This is a high level of residential mo- 45, 47]. Affective disorders (anxiety and depression) were bility , butisconsistentwithother studiesofthis the most common health conditions. Almost 8% of our population . On average, the cohort had more than cohort had a physician-diagnosed substance abuse dis- four ADGs, indicating they had a high level of comor- order; again consistent with other work showing a high bidity, which is consistent with the findings of other prevalence of drug and alcohol use among public housing researchers [7, 27, 28]. Previous studies reported that residents [4, 12–14]. All of these conditions were mea- public housing residents have a high prevalence of sured in the year before they moved into public housing. chronic physical health  and mental health condi- This suggests that public housing in Manitoba accepts tions [7, 9, 10, 13, 45, 46]. Theprevalenceof respira- and houses individuals with a high burden of disease - tory disease (which includes asthma, acute and chronic individuals who may have trouble obtaining and bronchitis, emphysema, and chronic airway obstruction) maintaining employment. Table 3 Unadjusted and adjusted estimates and 95% confidence intervals (CIs) for the period (post-move-in relative to pre-move-in) by month (30-day interval) interaction for healthcare use (N = 1942) Healthcare use Model Estimate 95% CI Hospitalization (OR) (all periods) Unadjusted 0.93 0.89, 0.97 Adjusted 0.92 0.88, 0.96 Hospitalization (OR) (admission period) Unadjusted 0.97 0.93, 1.01 Adjusted 0.97 0.93, 1.02 GP Visits (IRR) Unadjusted 0.99 0.98, 1.00 Adjusted 0.99 0.98, 1.00 Specialist Visits (OR) Unadjusted 1.00 1.00, 1.02 Adjusted 1.00 0.98, 1.02 Prescriptions (mean) Unadjusted −0.03 −0.04, −0.02 Adjusted −0.04 −0.05, −0.03 ED Visits (OR) Unadjusted 0.98 0.95, 1.01 Adjusted 0.98 0.95, 1.10 Note. Values in bold-face font are statistically significant at α = 0.05; Covariates in adjusted models = sex, age group, region of residence, income quintile, residential mobility, receipt of IA, physician-diagnosed mental and physical health conditions (i.e., schizophrenia, mood disorders, substance abuse disorders, injury, diabetes, respiratory illness, arthritis, cancer, hypertension), ADGs; Winnipeg residents only (N =960) Hinds et al. BMC Health Services Research (2018) 18:411 Page 9 of 13 Table 4 Unadjusted and adjusted estimates and 95% confidence intervals (CIs) for the average rate of change in each period for healthcare use (N = 1942) Healthcare use Period Unadjusted Adjusted Estimate 95% CI Estimate 95% CI Hospitalizations (OR) (all periods) Pre 1.06 1.03, 1.09 1.07 1.04, 1.10 Post 0.98 0.95, 1.01 0.98 0.95, 1.01 Hospitalizations (OR) (admission period only) Pre 1.04 1.01, 1.07 1.04 1.01, 1.07 Post 1.01 0.98, 1.04 1.01 0.98, 1.05 GP Visits (IRR) Pre 1.01 1.01, 1.02 1.01 1.01, 1.02 Post 1.00 1.00, 1.01 1.00 1.00, 1.01 Specialist Visits (OR) Pre 1.00 1.00, 1.02 1.00 0.99, 1.02 Post 1.00 0.99, 1.01 1.00 0.99, 1.01 Prescriptions (mean) Pre 0.05 0.04, 0.06 0.06 0.05, 0.07 Post 0.02 0.01, 0.03 0.02 0.01, 0.03 ED Visits (OR) Pre 1.02 1.01, 1.04 1.02 1.00, 1.04 Post 1.01 0.99, 1.02 1.00 0.99, 1.02 Note. Values in bold-face font are statistically significant at α = 0.05; Covariates in adjusted models were sex, age group, region of residence, income quintile, residential mobility, receipt of IA, physician-diagnosed mental and physical health conditions (i.e., schizophrenia, mood disorders, substance abuse disorders, injury, diabetes, respiratory illness, arthritis, cancer, hypertension), ADGs; Winnipeg residents only (N = 960) The cohort had a high use of healthcare services both fluctuated over the first year in public housing at levels before and after they moved into public housing; how- similar to the year before the move-in date. Our results ever, given the high burden of disease, the amount of are consistent with Smith, Alexander, and Easterlow healthcare use is not surprising. In any 30-day interval, (1997) . They found that healthcare use changed for we found that approximately 40% had a GP visit, 12% individuals who moved into medical priority public had a specialist visits, 8% (of Winnipeg residents) visited housing (a practice in Britain of prioritizing individuals an ED, and 2.5% were hospitalized. There was also evi- with health or mobility problems to receive social hous- dence that healthcare use changed when individuals ing) . Most people reported their healthcare use de- moved into public housing, but the direction of the creased; specifically, one in five people visited their change varied by the type of health service. We hypothe- family doctor less and had fewer outpatient visits, one in sized that health may be associated with applying to four people had fewer consultant/specialist visits, and public housing and hence healthcare use may increase one in three people spent less time in hospital. prior to tenancy in public housing, and in fact, 8.7% of There was evidence that the average rate of change the cohort reported a health/medical reason as their in the odds/rate of being hospitalized/visiting a GP in motive for applying to public housing. Healthcare use the pre-move-in period was higher than in the post- did increase up to approximately three months prior to move-in period and that the odds/rate in the pre- the move-in date and then within a few months after the move-in period increased over time, while it did not move-in date, the percentage of the cohort using all increase in the post-move-in period. Additionally, forms of healthcare services decreased, except for those there was evidence that the average rate of change in using prescription medications. the mean use of prescriptions drugs increased over Approximately 50% of cohort members applied to time in both periods, but it increased more quickly in public housing three months before their move-in date. the pre-move-in period. These findings may suggest Further research is needed to determine whether there is that public housing interrupted the need for some an association between application approval (i.e., ap- types of healthcare and that once individuals were proved/not approved, length of time to be approved) housed in public housing they better adhered to their and health, and between application reason (i.e., health/ regular source of care. non-health) and length of time to be housed. Wood et al. adopted a similar methodology of using Prescription use steadily increased over the two-year linked administrative data to compare healthcare use be- time period. Specialist visits and hospitalizations were fore and after people who were homeless moved into maintained at a level lower in the post-move-in period public housing; however, they used one year intervals in- compared to the pre-move-in date period. The percent- stead of 30-day intervals. They found that access and age of the cohort who visited GPs and emergency rooms frequency of some forms of healthcare use in the first Hinds et al. BMC Health Services Research (2018) 18:411 Page 10 of 13 year in public housing decreased compared to the Approximately 63% of social housing units in the year before they moved in . Specifically, there was province are not directly managed by the government, a significant decrease in access to emergency depart- but are operated by cooperatives, non-profit groups, ments, overnight hospital stays, admissions to the in- and property management agencies. Residents of these tensive care unit (ICU), receipt of psychiatric care, forms of social housing were not included as there receipt of mental health services, and use of three was no individual-level administrative data available. prescriptions (i.e., Methadone, Subutex, Suboxone). Also, we excluded individuals who resided in public Also, among the individuals who visited emergency housing for less than a year and this may have re- departments, there was little difference in the mean sulted in some selection bias. We suspect that people number of visits between the two periods; however, who have short stays in public housing (i.e., less than there was a reduction in the mean number of days in one year), are generally less healthy (i.e., more mental hospital, ICU days, days admitted for psychiatric care, health issues and substance use issues) and are more mean number of hours of receipt of mental health frequent users of healthcare services; therefore, we services, and mean number of prescriptions in the may have found stronger effects had we included year after the move-in date compared with the year them in the cohort. before. Interestingly, the decrease in healthcare use Measurement error maybe associated with some of was most pronounced for individuals living in public the covariates. For example, the diagnoses of the housing between one and four years. However, when health conditions are based on physician visits and Wood et al.  compared the average healthcare use hospitalizations. Only one diagnosiscodeisrecorded in the three year period before the public housing for each physician visit. Consequently, the number of move-in date with healthcare use in the one year people with any of the health conditions may be period after the move-in date, some changes in the underestimated. Residential mobility may have been magnitude of healthcare use were found. Specifically, underestimated if address changes were not reported overnight hospital stays and use of mental health to the province. However, given the high level of services were more common in the year after the healthcare utilization by the cohort, and that fact move-indateand the changeis the useofpsychiatric that healthcare providers and hospitals require pa- services varied by program status. tients to have up-to-date information on their health card, any underestimation in residential mobility is Study strengths likely minimal. Our study has a number of strengths. One of the We did not include a comparison group; conse- strengths is that we used population-based administra- quently we cannot determine whether changes in tive data. The data is owned and managed by provincial healthcare use were just reflective of changes in use by government departments for administrative purposes the larger population. We plan to conduct a follow-up (e.g., physician reimbursement) and thus are of high study using the administrative data to compare health- quality with no missing information in the main fields. care use over time between residents of public housing The only ‘missing’ data is the ‘not found’ category for and a comparison group matched from the general income quintile, which affected 1.2% of the cohort. population as well as a group who applied to public These data are missing because postal codes were not housing but did not move-in (i.e., were not approved able to be assigned to a DA or a DA had a small non- for public housing or canceled their application to pub- institutionalized population. lic housing). We were unable to determine what con- We linked public housing data to health data at an tributed to changes in healthcare use. Potential individual-level to comprehensively examine healthcare contributing factors to the decreased use include better before and after the public housing move-in date. Wood access to informal caregivers [48, 50], better housing et al. compared healthcare use as two periods (pre and , improved access to social services, including post-move-in date) . We divided these periods into family resource centres , more income to spend on twelve 30-day intervals. These shorter units of time nutritious food and recreational activities (income ef- allowed us to examine trends, providing evidence there fect) , and increased access to other services . A may be health factors precipitating an application to future qualitative research study might shed light on public housing as well as evidence of a transition or ad- the reasons for changes in use. While the drop in justment period to public housing. healthcare use may reflect decreased access to health services, this is unlikely as others have found that health Study limitations and future directions services are located close to public housing [53, 54]. Re- Our cohort was limited to residents of public hous- searchers found that residing in a socioeconomically dis- ing, housing that is directly managed by the province. advantaged neighbourhood is associated with decreased Hinds et al. BMC Health Services Research (2018) 18:411 Page 11 of 13 T5 healthcare access, even after controlling for healthcare Appendix supply and individual-level characteristics . A fu- Table 5 ICD-9-CM and ICD-10-CA codes for selected health ture study is warranted to determine whether this is conditions true in Winnipeg and in Manitoba. The public housing Condition ICD-9-CM ICD-10-CA application asks individuals where they want live, but Physical it is not known how closely placement matches Respiratory Illness 466, 490–493, 496 J20, J21, J40 – J45 preferences. A future study could determine the dis- Diabetes 250 E10 – E14 tance individuals move when they are placed in public Hypertension 401–405 I10 – I15 housing, describe residential mobility patterns (i.e., are Cancer 14–20 C00 – C97 individuals placed in the same neighbourhood as on their application), and determine whether distance to Arthritis 274, 446, 710–721, M00 – M03, M05 – M07, 725–729, 739 M10 – M25, M30 – M36, healthcare services (i.e., hospitals, EDs, primary health- M65 – M79 care provider) changes after a move and whether this Injury 80–99 S00 – S99, T00 – T98 varies by health region. Additionally, to move within Mental Disorder Manitoba Housing, individuals have to reapply; thus, it would is possible to examine residential mobility pat- Schizophrenia 295 F20, F21, F25, F232 terns of public housing residents, and determine how Affective Disorder 296, 300, 309, 311 F31 – F33, F40 – F42, F44, this is related to their pre-public housing location of F48, F99, F341, F380, F381, F410, F411, F412, F413, F418, residence and access to healthcare. Housing adminis- F419, F431, F432, F438, F450, trators would likely find it useful to know where F451, F452, F530, F680, F930 people (with health conditions) are being housed in re- Substance Abuse 291, 292, 303–305 F10 – F19, F55 lation to their healthcare providers and their prior so- Disorder cial network. Additionally, since Wood et al. found varying the length of the pre and post move-in periods (one and three years) affected the findings , a follow-up study could examine healthcare trends over Additional files longer periods. Lastly, further research could examine changes in use by specialist type, changes in physician Additional file 1: Table S1. Unadjusted Point Estimates and 95% Confidence Intervals (CI) for the Healthcare Utilization Measures over visits by the reason for the visit (e.g., mental health, Time. (DOCX 19 kb) physical health, and preventive health (i.e., medical Additional file 2: Table S2. Chi-Square Test Statistics and P-values for screening) reasons) as well as changes in healthcare use Main and Interaction Effects. (DOCX 17 kb) for people with different health conditions. Additional file 3: Table S3. Model Estimates and 95% Confidence Intervals (CIs). (DOCX 23 kb) Conclusion In summary, the use of several types of healthcare ser- Abbreviations ACG: Adjusted Clinical Group; ADG: Aggregated Diagnostic Groups; vices (i.e., specialist visits, hospitalizations) declined after ATC: Anatomical Therapeutic Chemical; CI: Confidence interval; people moved into public housing. However, for some DA: Dissemination area; DPIN: Drug Program Information Network; forms of healthcare services (GP visits, emergency de- ED: Emergency department; GEE: Generalized estimating equations; GP: General practitioner; HOPE: Housing Opportunities for People partment visits), the decrease in use was only observed Everywhere; IA: Income assistance; ICD-10-CA: International Classification of for the first few months after the move-in date; percent- Diseases, 10th revision, Canadian version; ICD-9-CM: International ages rebounded shortly thereafter. This rebounding ef- Classification of Diseases, 9th revision, Clinical Modification; ICU: Intensive care unit; IQ: Income quintile; IRR: Incident rate ratio; NF: Not found; fect could be further examined to understand why this OR: Odds ratio; Q1: Quartile 1 (25th percentile); Q3: Quartile 3 (75th occurred. percentile); QIC: Quasi-likelihood information criterion; SD: Standard In general, since public housing residents are high deviation; TMS: Tenant Management System users of healthcare services and tend to experience a Acknowledgements high burden of disease, a need exists to strategically Thanks are owed to Heather Prior for data extraction from the locate health and social services in public housing de- Manitoba Centre for Health Policy Population Research Data Repository. Thanks are also owed to Kristine Kroeker for providing statistical velopments, preferably using an integrative, commu- consulting advice. nity/client-centred approach, such that there are a range of services in one location (e.g., Community Funding Health Centres or ACCESS Centres) tailored to the This work was supported by funding to AMH from Research Manitoba (PhD Dissertation Award). LML was supported by a Manitoba Research Chair from community’sneeds. As May recommends, housing Research Manitoba. Research Manitoba did not participate in the design of policy needs to be linked with social policy for service the study or the collection, analysis, or interpretation of the data or in integration . writing the manuscript. Hinds et al. BMC Health Services Research (2018) 18:411 Page 12 of 13 Availability of data and materials 5. Buchner D, Nicola R, Martin M, Patrick D. Physical activity and health Data used in this article was derived from administrative health and social promotion for older adults in public housing. Am J Prev Med. 1997; data as a secondary use. The data was provided under specific data sharing 13(Suppl 2):57–62. agreements only for approved use at Manitoba Centre for Health Policy 6. Lambertino A, Turyk ME, Curtis L, Persky VW. Asthma morbidity in adult (MCHP). The original source data is not owned by the researchers or MCHP Chicago public housing residents. J Asthma. 2009;46:202–6. and as such cannot be provided to a public repository. The original data 7. Bazargan M, Calderon JL, Heslin KC, Mentes C, Shaheen MA, Ahdout J, source and approval for use has been noted in the Ethics Approval section Baker RS. A profile of chronic mental and physical conditions among of the article. Where necessary, source data specific to this article or project African-American and Latino children in urban public housing. Ethn Dis. may be reviewed at MCHP with the consent of the original data providers, 2005;15(Supplement):S5-3–9. along with the required privacy and ethical review bodies. 8. Smith M, Finlayson G, Martens P, Dunn J, Prior H, Taylor C, Soodeen R-A, Burchill C, Guenette W, Hinds A. Social housing in Manitoba part II: social housing and health in Manitoba: a first look. Winnipeg, MB: Manitoba Disclaimer Centre for Health Policy; 2013. The authors acknowledge the Manitoba Centre for Health Policy for use of 9. Simning A, Wijngaarden E, Van CY. Anxiety, mood, and substance use data contained in the Population Research Data Repository under project disorders in United States African-American public housing residents. Soc #2015–002 (HIPC#2014/2015.29). The results and conclusions are those of the Psychiatry Psychiatr Epidemiol. 2011;46:983–92. authors and no official endorsement by the Manitoba Centre for Health 10. Rabins PV, Black B, German P, Roca R, McGuire M, Brant L, Cook J. The Policy, Manitoba Health, or other data providers is intended or should be prevalence of psychiatric disorders in elderly residents of public housing. inferred. Data used in this study are from the Population Research Data Journals Gerontol Med Sci. 1996;51A:M319–24. Repository housed at the Manitoba Centre for Health Policy, University of 11. Gonyea JG, Bachman SS. Correlates of psychological distress among older Manitoba and were derived from data provided by Manitoba Health, the residents of urban public housing. J Loss Trauma. 2008;13:243–60. Winnipeg Regional Health Authority, and the Provincial Government 12. Williams CD, Adams-Campbell LL. Addictive behaviors and depression Department of Families. among African Americans residing in a public housing community. Addict Behav. 2000;25:45–56. Authors’ contributions 13. Cummings SM, Cooper RL, Johnson C. Alcohol misuse among older adult AMH, LML, BB, JD, and LLR contributed to the study conception, design, public housing residents. J Gerontol Soc Work. 2013;56:407–22. interpretation of the data, drafting and critical revision of the manuscript. 14. DeKeseredy WS, Schwartz MD, Alvi S, Tomaszewski EA. Crime victimization, AMH contributed to the data analysis, interpretation of the data, and writing alcohol consumption, and drug use in Canadian public housing. J Crim and revision of the manuscript. All authors participated in editing and Justice. 2003;31:383–96. revising the manuscript. All authors read and approved the final manuscript. 15. Yu M, Nebbitt VE, Lombe M, Pitner RO, Salas-Wright CP. Understanding tobacco use among urban African American adolescents living in public Ethics approval and consent to participate housing communities: a test of problem behavior theory. Addict Behav. Ethics approval was obtained from the University of Manitoba Health Research 2012;37:978–81. Ethics Board (H2014:381). Data access was approved by the Manitoba Health 16. Sikkema KJ, Kelly JA, Winett RA, Solomon LJ, Cargill VA, Roffman RA, Information Privacy Committee (2014/15–29), the Winnipeg Regional Health McAuliffe TL, Heckman TG, Anderson EA, Wagstaff DA, Norman AD, Perry Authority, and the Provincial Government Department of Families. This study used MJ, Crumble DA, Mercer MB. Outcomes of a randomized community-level population-based administrative data. Due to the retrospective nature of the study HIV prevention intervention for women living in 18 low-income housing design and the large cohort size and geographical distribution of the study developments. Am J Public Health. 2000;90:57–63. members, it is not feasible to obtain consent to participate or consent to publish. 17. Sikkema KJ, Koob JJ, Cargill VC, Kelly JA, Desiderato LL, Roffman RA, Norman AD, Shabazz M, Copeland C, Winett RA, Steiner S, Lemke AL. Levels and Competing interests predictors of HIV risk behavior among women in low-income public The authors declare that they have no competing interests. housing developments. Public Health Rep. 1995;110:707–13. 18. Roman CG, Knight CR, Chalfin A, Popkin SJ. The relation of the perceived environment to fear, physical activity, and health in public housing Publisher’sNote developments: evidence from Chicago. J Public Health Policy. 2009; Springer Nature remains neutral with regard to jurisdictional claims in 30(SUPPL. 1):286–308. published maps and institutional affiliations. 19. Lewis CE, Raczynski JM, Heath GW, Levinson R, Cutter GR. Physical activity of public housing residents in Birmingham, Alabama. Am J Public Heal. 1993; Author details 83:1016–20. Department of Community Health Sciences, University of Manitoba, 20. Bennett GG, McNeill LH, Wolin KY, Duncan DT, Puleo E, Emmons KM. Safe to S113-750 Bannatyne Ave, Winnipeg, Manitoba R3E 0W3, Canada. Cross walk? Neighborhood safety and physical activity among public housing Ministry and Community Partnership Initiatives Community and Social residents. PLoS Med. 2007;4:1599–607. Services, 3rd floor, 10044-108 Street, Edmonton, Alberta T5J 5E6, Canada. 21. Lee RE, Mama SK, McAlexander KP, Adamus H, Medina AV. Neighborhood Department of Geography, University of Winnipeg, 515 Portage Avenue, and PA: neighborhood factors and physical activity in African American Winnipeg, Manitoba R3B 2E9, Canada. public housing residents. J Phys Act Health. 2011;8 Suppl 1(Suppl 1):S83–90. 22. Shelton RC, McNeill LH, Puleo E, Wolin KY, Emmons KM, Bennett GG. The Received: 18 July 2017 Accepted: 11 April 2018 association between social factors and physical activity among low-income adults living in public housing. Am J Public Health. 2011;101:2102–10. 23. Scammell MK, Torres S, Wayman J, Greenwood N, Thomas G, Kozlowski L, Bowen D. Balancing act: approaches to healthy eating and physical activity References among Boston public housing residents. J Prev Interv Community. 1. Government Assisted Housing [https://www.cmhc-schl.gc.ca/en/inpr/ 2015;43:109–22. afhoce/afhoce_021.cfm]. Accessed 20 Apr 2018. 24. Heinrich K, Lee RE, Suminski RR, Regan GR, Reese-Smith JY, Howard HH, 2. Dockery AM, Ong R, Whelan S, Wood G. The relationship between public Haddock CK, Carlos Poston WS, Ahluwalia JS. Associations between the built housing wait lists, public housing tenure and labour market outcomes. environment and physical activity in public housing residents. Int J Behav Melbourne: Australian Housing and Urban Research Institute; 2008. Nutr Phys Act. 2007;4:56. 3. Digenis-Bury EC, Brooks DR, Chen L, Ostrem M, Horsburgh CR. Use of a population-based survey to describe the health of Boston public housing 25. Hinds AM, Bechtel B, Distasio J, Roos LL, Lix LM. Health and social predictors residents. Am J Public Health. 2008;98:85–91. of applications to public housing: a population-based analysis. J Epidemiol 4. Wiggers J, Radvan D, Clover K, Hazell T, Alexander J, Considine R. Public Community Health. 2016;70:1229–35. housing, public health: health needs of public housing tenants. Aust N Z J 26. McNeill LH, Coeling M, Puleo E, Suarez EG, Bennett GG, Emmons KM. Public Health. 2001;25:111–4. Colorectal cancer prevention for low-income, sociodemographically-diverse Hinds et al. BMC Health Services Research (2018) 18:411 Page 13 of 13 adults in public housing: baseline findings of a randomized controlled trial. 49. Wood L, Flatau P, Zaretzky K, Foster S, Vallesi S, Miscenko D. What are the BMC Public Health. 2009;9:1–15. health, social and economic benefits of providing public housing and 27. Black BS, Rabins PV, German P, McGuire M, Roca R. Need and unmet need support to formerly homeless people? Melbourne: Australian Housing and for mental health care among elderly public housing residents. Urban Research Institute; 2016. Gerontologist. 1997;37:717–28. 50. May J. Social lives in social housing: resident connections to social services. Ottawa, ON: Canadian Policy Research Networks; 2007. 28. Black BS, Rabins PV, German P, Roca R, McGuire M, Brant LJ. Use of formal 51. Fertig A, Reingold D. Public housing, health, and health behaviors: is there a and informal sources of mental health care among older African-American connection? J Policy Anal Manag. 2007;26:831–59. public-housing residents. Psychol Med. 1998;28:519–30. 52. Larios L, Lyle M, Paille M, Smith M: Investigating access to health and social 29. Kersten EE, LeWinn KZ, Gottlieb L, Jutte DP, Adler NE. San Francisco children Services for Residents of Manitoba housing. 2013. living in redeveloped public housing used acute services less than children 53. Apparicio P, Séguin AM. Measuring the accessibility of services and facilities in older public housing. Health Aff. 2014;33:2230–7. for residents of public housing in Montréal. Urban Stud. 2006;43:187–211. 30. Finlayson G, Smith M, Burchill C, Towns D, Peeler W, Soodeen RA, Prior H, 54. Cooper HL, Wodarski S, Cummings J, Hunter-Jones J, Karnes C, Ross Z, Druss Huq S, Guenette W: Social housing in Manitoba part I: Mantioba social B, Bonney LE. Public housing relocations in Atlanta, Georgia, and declines in housing data. 2013. spatial access to safety net primary care. Health Place. 2012;18:1255–60. 31. Martens PJ, Chateau DG, Burland EMJ, Finlayson GS, Smith MJ, Taylor CR, 55. Kirby JB, Kaneda T. Neighborhood socioeconomic disadvantage and access Brownell MD, Nickel NC, Katz A, Bolton JM. The effect of neighborhood to health care. J Health Soc Behav. 2005;46:15–31. socioeconomic status on education and health outcomes for children living in social housing. Am J Public Health. 2014;104:2103–13. 32. Watson DE, Katz A, Reid RJ, Bogdanovic B, Roos N, Heppner P. Family physician workloads and access to care in Winnipeg: 1991 to 2001. Can Med Assoc J. 2004;171:339–42. 33. Lix LM, Yao X, Kephart G, Quan H, Smith M, Kuwornu JP, Manoharan N, Kouokam W, Sikdar K. A prediction model to estimate completeness of electronic physician claims databases. BMJ Open. 2015;5:e006858. 34. Lix LM, Kuwornu JP, Kroeker K, Kephart G, Sikdar KC, Smith M, Quan H. Estimating the completeness of physician billing claims for diabetes case ascertainment using population-based prescription drug data. Heal Promot chronic Dis Prev Canada Res Policy Pract. 2016;36:54–60. 35. Doupe M, Chateau D, Derksen S, Sarkar J, Lobato de Faria R, Strome T, Soodeen R-A, McCulloch S, Dahl M. Factors affecting emergency department waiting room times in Winnipeg. Winnipeg, MB: Manitoba Centre for Health Policy; 2017. 36. Lix LM, Hinds A, DeVerteuil G, Robinson JR, Walker J, Roos LL. Residential mobility and severe mental illness: a population-based analysis. Adm Policy Ment Heal Ment Heal Serv Res. 2006;33:160–71. 37. DeVerteuil G, Hinds A, Lix L, Walker J, Robinson R, Roos LL. Mental health and the city: intra-urban mobility among individuals with schizophrenia. Heal Place. 2007;13:310–23. 38. Heaman M, Kingston D, Helewa ME, Brownell M, Derksen S, Bogdanovic B, Mcgowan K, Bailly A. Perinatal services and outcomes in Manitoba. Winnipeg, MB; 2012. 39. Roos NP, Mustard CA. Variation in health and health care use by socioeconomic status in Winnipeg, Canada: does the system work well? Yes and no. Milbank Q. 1997;75:89–111. 40. Roos LL, Walld R, Witt J. Adolescent outcomes and opportunities in a Canadian province: looking at siblings and neighbors. BMC Public Health. 2014;14:1–16. 41. Austin PC, Van Walraven C, Wodchis WP, Newman A, Anderson GM. Using the Johns Hopkins aggregated diagnosis groups (ADGs) to predict mortality in a general adult population cohort in Ontario, Canada. Med Care. 2011;49:932–9. 42. Lix LM, Quail J, Fadahunsi O, Teare GF. Predictive performance of comorbidity measures in administrative databases for diabetes cohorts. BMC Health Serv Res. 2013;13:1. 43. Martens P, Fransoo R, The Need to Know Team, Burland E, Prior H, Burchill C, Romphf L, Chateau D, Bailly A, Ouelette C. What works? A first look at evaluating Manitoba’s regional health programs and policies at a population level. Winnipeg, MB: Manitoba Centre for Health Policy; 2008. 44. SAS Institute Inc. The GENMOD procedure. In SAS/STAT 92 User’s guide. Cary, NC: SAS Institute Inc; 2013. p. 2871–3076. 45. Pollack CE, Green HD, Kennedy DP, Griffin BA, Kennedy-Hendricks A, Burkhauser S, Schwartz H. The impact of public housing on social networks: a natural experiment. Am J Public Health. 2014;104:1642–9. 46. Simning A, Van Wijngaarden E, Fisher SG, Richardson TM, Conwell Y. Mental healthcare need and service utilization in older adults living in public housing. Am J Geriatr Psychiatry. 2012;20:441–51. 47. Manfredi C, Lacey L, Wamecke R, Buis M. Smoking-related behaviors, beliefs, and social environment of young black women in subsidized public housing in Chicago. Am J Public Health. 1992;82:267–73. 48. Smith SJ, Alexander A, Easterlow D. Rehousing as a health intervention: miracle or mirage? Heal Place. 1997;3:203–16.
BMC Health Services Research – Springer Journals
Published: Jun 5, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
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.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera