Retention in care and mortality trends among patients receiving comprehensive care for HIV infection: a retrospective cohort study
==================================================================================================================================

* Ana C. Ulloa
* Cathy Puskas
* Benita Yip
* Wendy Zhang
* Cole Stanley
* Sarah Stone
* Miguel Pedromingo
* Viviane Dias Lima
* Julio S.G. Montaner
* Silvia Guillemi
* Rolando Barrios

## Abstract

**Background:** Studies examining the relation between comprehensive care and health outcomes associated with comorbidities unrelated to HIV infection have focused mainly on the health outcomes of HIV-infected people and comorbid substance use disorders. We aimed to assess the impact of retention in comprehensive HIV infection care on overall, AIDS-related and non–AIDS-related mortality.

**Methods:** Using a retrospective cohort design, we collected data for HIV-infected patients aged 19 years or more who first visited a comprehensive HIV infection clinic in Vancouver between Jan. 1, 2004, and Dec. 31, 2014. We defined retention in care as visit constancy (whether the patient attended the clinic at least once per given period) of 75% or greater. We used Poisson regression modelling to examine mortality trends. We performed Cox proportional hazards modelling to assess survival by retention during the first year of follow-up and identify factors associated with death.

**Results:** A total of 2101 patients were included in the study. Of the 2101, 1340 (63.8%) were retained in the first year of care, and 271 (12.9%) died during the study period. Among the 264 cases in which the cause of death was known, although the primary underlying cause of death (74 [28.0%]) was AIDS-related, half of all AIDS-related deaths (37/74 [50%]) occurred early in the study (2004–2007). In later years, most deaths (147/184 [79.9%]) were non–AIDS-related. Overall mortality was significantly reduced among patients with higher retention in care during the first year of follow-up (per 20% increase in visit constancy; adjusted hazard ratio [HR] 0.87, 95% confidence interval [CI] 0.79–0.96). Higher retention was also associated with reduced risk of AIDS-related death (adjusted HR 0.79, 95% CI 0.64–0.97).

**Interpretation:** Although there was an overall trend toward decreased AIDS-related mortality over time, retention in care markedly decreased the likelihood of death. Maintaining patient engagement in comprehensive ancillary care is a patient-centred way of decreasing mortality rates among HIV-infected people.

Thanks to modern antiretroviral therapy (ART), life expectancy for people with HIV infection has improved substantially in the last decade.1 Comprehensive care models have been shown to improve health outcomes related to HIV infection, increasing adherence to HIV infection care and rates of viral load suppression.2,3 Few studies have examined the relation between comprehensive care and health outcomes associated with comorbidities unrelated to HIV infection; these studies have focused mainly on the health outcomes of HIV-infected people and comorbid substance use disorders.4–6

At the end of 2014, there were about 75 500 people with HIV infection in Canada, with 16.0% (12 100) residing in British Columbia.7 The John Ruedy Immunodeficiency Clinic at St. Paul’s Hospital, Vancouver, was established in 2003 to address the needs of HIV-infected people without primary care or specialist access. At the end of December 2015, the clinic was caring for 1306 active patients (roughly 10% of those with HIV infection in BC). The clinic serves a diverse community of HIV-infected people and provides comprehensive and supportive HIV infection care. The primary objective of this study was to examine trends in all-cause and cause-specific mortality among patients enrolled in a comprehensive care HIV infection clinic. In particular, we wished to identify the impact of retention in care on overall, AIDS-related and non–AIDS-related mortality to better understand the relation between retention in comprehensive care HIV infection programs and improvement in overall health outcomes among people with HIV infection.

## Methods

### Study design and population

The John Ruedy Immunodeficiency Clinic is an interdisciplinary outpatient HIV clinic providing low-barrier access to comprehensive care. The clinic is staffed by an interdisciplinary team of family physicians, specialists, nurses, pharmacists, counsellors, social workers, peer navigators and nutritionists who provide numerous services to HIV-infected patients, including access to addiction services, an on-site pharmacy, rapid referral to HIV infection care and specialists treating infectious and noninfectious comorbidities, and rapid access to social work support and counselling. In BC, people infected with HIV have universal and free access to ART. Since 2011, ART initiation has been recommended to all those with HIV infection regardless of CD4 count.8

This retrospective cohort study included HIV-infected people aged 19 years or more enrolled in the clinic who first visited the clinic between Jan. 1, 2004, and Dec. 31, 2014. Inclusion in the study required at least 1 recorded visit to the clinic and at least 1 documented CD4 count and plasma viral load test during the study period. All patients were followed until the earliest of death, migration or study censor date (Dec. 31, 2015).

### Data collection

We obtained data for the participants from the Drug Treatment Program, the clinic’s internal database and BC Vital Statistics. The Drug Treatment Program, developed and managed by the BC Centre for Excellence in HIV/AIDS, collects demographic and HIV-related clinical, laboratory and treatment-related information for all people with HIV infection who have received ART in BC. Data collected by the program have been previously described.9,10 Detailed clinical and demographic characteristics of enrolled patients were extracted from the clinic’s internal database. For deceased patients, a trained research assistant used a standardized form to collect information on comorbidities, opportunistic infections, death and autopsy results (if available) from the hospital electronic health records. We obtained cause of death from BC’s Vital Statistics Agency.

### Primary outcomes

The primary outcome variables were overall and cause-specific mortality as determined from physician reviews based on the Coding Causes of Death in HIV (CoDe) protocol,11 a widely used uniform classification system that has been previously evaluated.12 Using CoDe, 2 of C.S., S.S. or M.P. independently reviewed each patient’s deidentified information to ascertain cause of death. Two reviewers received a batch of 15 cases at a time; these were rotated so that each of the 3 reviewers was matched equally with the other 2 reviewers. To address any potential discrepancies, the reviewers attended an initial training session where a practice batch was given. In the event of disagreement, a third physician (R.B. or S.G.) served as a tiebreaker.

Underlying causes of death were classified into 3 categories: AIDS-related, non–AIDS-related and unknown/unclassifiable. They were then subdivided according to CoDe protocol categories, which include 18 specific code classes, 6 general classifications and 2 codes for unclassifiable causes.12 The final list of subcategories used in this study included the following specific code classifications: AIDS-related (AIDS-related infection, AIDS malignancy and AIDS [ongoing active disease]), malignancy (other than AIDS- or hepatitis-related), substance use (active), chronic viral hepatitis (progression of/complication to), infection (other than AIDS-related), chronic obstructive pulmonary disease, myocardial infarction or other ischemic heart disease (general classifications: heart or vascular disease), psychiatric disease, and unclassifiable causes: other and unknown/unclassifiable causes. All causes of death apart from unknown/unclassifiable with a frequency less than 7 were grouped together as “other.”

### Covariates

We collected data for all patients on sex (male/female), age, risk factors for HIV infection (injection drug use, men who have sex with men [MSM], injection drug use and MSM, other [heterosexual risk of HIV infection alone or in combination with injection drug use or MSM, or accidental exposure], unknown), age at clinic enrolment, year of clinic enrolment, receipt of ART (ever during study period) (yes/no), baseline (most recent within 6 mo of first clinic visit) and most recent CD4 count, baseline and most recent HIV plasma viral load, retention in care during the first year of follow-up (yes/no) and ART adherence during the first year of follow-up (yes/no). For deceased patients, additional information was collected on chronic hepatitis B status (yes/no) and chronic hepatitis C status (yes/no), and the presence of AIDS-related illnesses in the year before death (yes/no).

We measured retention in care using visit constancy, or whether the patient attended the clinic at least once per given period.13,14 We divided the number of intervals in which a patient had at least 1 primary or specialized care clinic visit by the total number of intervals during follow-up and reported the result as a percentage. We followed Adult Therapeutic Guidelines recommendations for laboratory and clinical monitoring for 2004–201515 and clinical expertise to determine visit constancy intervals. Twelve-month observation periods were broken down into four 3-month intervals (2004–2008), three 4-month intervals (2009–2010) and two 6-month intervals (2011–2015). Patients with a visit constancy of 75% or greater were classified as retained in care. Consultation with clinic physicians suggested that this value best indicated retention in care, as many patients could not realistically attain 100% visit constancy. Visit constancy was also expressed in 20% intervals for consistency with the literature.13

We used pharmacy refill records (ART dispensation dates logged by the Drug Treatment Program) to assess ART adherence; this measure of adherence has been previously validated. 16 We divided the number of days for which ART was dispensed by the total number of days of follow-up during the first year to obtain the proportion of time for which the patient had ART coverage.17 We defined optimal ART adherence as 95% or more; this cut-off has been consistently associated with virologic suppression18,19 and decreased likelihood of resistance.20

### Statistical analysis

We calculated all-cause and cause-specific mortality rates per 1000 person-years and used a Poisson regression model to examine trends for overall, AIDS-related and non–AIDS-related mortality.21 We assessed agreement between physician reviews for cause of death using the κ statistic.22 We compared characteristics of living and deceased patients, and between patients who died from AIDS-related and non–AIDS-related causes, using χ2 statistics for categorical variables and the Wilcoxon rank sum test for continuous variables. We assessed survival using Cox proportional hazards models for overall, AIDS-related and non–AIDS-related mortality, stratified by retention during the first year of follow-up. We used unadjusted and adjusted Cox proportional hazards models23 to identify factors independently associated with increased risk of death. We did not use ART adherence as a variable in the multivariate models, as it was found to be collinear with visit constancy. For the multivariate model, we used the Akaike information criterion for model selection.24 We used the Supremum test to assess the proportionality assumption and nonsignificant *p* values. In a sensitivity analysis, we considered competing risks between AIDS-related and non–AIDS-related deaths, but model results were very similar (data not shown). All analyses were conducted with SAS software v9.4 (SAS Institute) with a level of significance set at 0.05.

### Ethics approval

The study protocol received ethics approval from the University of British Columbia Providence Health Care Research Institute Research Ethics Board.

## Results

There were 2123 people with HIV infection enrolled in the clinic with a first visit between Jan. 1, 2004, and Dec. 31, 2014; of these, 22 did not meet our inclusion criteria. Thus, 2101 patients were included in the study. They accounted for 11 954 person-years (median 5 yr [25th, 75th percentiles 2, 9]) of follow-up. Table 1 presents the patients’ overall characteristics, stratified by death during the study period. Patients who remained alive were more likely than those who died to be MSM, to be younger at first visit, to have a higher CD4 count and lower plasma viral load at baseline and last visit, and to be retained in care and adherent to ART during the first year of follow-up (*p* < 0.001).

View this table:
[Table 1:](http://www.cmajopen.ca/content/7/2/E236/T1)

Table 1: 
Descriptive characteristics of patients enrolled in the John Ruedy Immunodeficiency

### Mortality rate

Of the 2101 patients, 271 (12.9%) died during the study period. The overall crude mortality rate was 22.7 per 1000 person-years (95% confidence interval [CI] 20.0–25.5). Specific causes of death were classifiable for 264 cases (97.4%). After we removed the 7 cases with unknown/unclassifiable cause of death, interrater reliability for cause of death was 89% (95% CI 85%–93%), with a κ coefficient of 0.73 (*p* < 0.001).

Poisson regression analysis results for all classifiable causes of death are shown in Figure 1. There were no AIDS-related deaths recorded in 2004; the AIDS-related mortality rate decreased from 36.2 per 1000 person-years in 2005 to 2.6 per 1000 person-years in 2015 (*p* < 0.001). The non–AIDS-related mortality rate decreased from 37.4 per 1000 person-years in 2004 to 14.4 per 1000 person-years in 2015 (*p* = 0.002). Table 2 shows detailed classifications of underlying causes of death and crude mortality rates per 1000 person-years.

![Figure 1:](http://www.cmajopen.ca/https://www.cmajopen.ca/content/cmajo/7/2/E236/F1.medium.gif)

[Figure 1:](http://www.cmajopen.ca/content/7/2/E236/F1)

Figure 1: 
Crude rates for overall, AIDS-related and non–AIDS-related mortality per 1000 person-years over time from 2005 to 2015. The year 2004 has been omitted from analysis as there were no AIDS-related deaths in that year. Patients with unknown cause of death (*n* = 7) were excluded.

View this table:
[Table 2:](http://www.cmajopen.ca/content/7/2/E236/T2)

Table 2: 
Underlying cause of death and crude mortality rates per 1000 person-years for the clinic cohort, 2004–2015 (11 954 total person-years)

### Characteristics associated with death and survival

Among the 264 patients for whom specific cause of death was classifiable, relative to patients who died from non–AIDS-related causes, those who died from AIDS-related causes were younger at first visit (*p* = 0.005) and at death (*p* < 0.001), had a lower CD4 count and higher plasma viral load at baseline and at death (*p* < 0.001), were less likely to be ART adherent in the first year of follow-up (*p* = 0.007) and had more AIDS-related illnesses in the year before death (*p* < 0.001) (Table 3).

View this table:
[Table 3:](http://www.cmajopen.ca/content/7/2/E236/T3)

Table 3: 
Descriptive characteristics of deceased patients, 2004–2015

The probability of death from all causes decreased from 15% to 7% as visit constancy increased (*p* = 0.001) (Figure 2); a decrease was observed for both AIDS-related (5% to 2%, *p* = 0.03) and non–AIDS-related (12% to 6%, *p* = 0.07) mortality. As 1 year of follow-up is required to assess visit constancy, all events occurring in the year that the patient enrolled in the clinic could not be included in this analysis. Consequently, the sample size was reduced to 190 deceased patients.

![Figure 2:](http://www.cmajopen.ca/https://www.cmajopen.ca/content/cmajo/7/2/E236/F2.medium.gif)

[Figure 2:](http://www.cmajopen.ca/content/7/2/E236/F2)

Figure 2: 
Probability of overall, AIDS-related, and non–AIDS-related mortality by level of visit constancy during the first year of follow-up (190 deceased patients) (Cox proportional hazards model). Patients who did not have data for first-year visit constancy or baseline CD4 count/plasma viral load were excluded from the model (*n* = 120).

After we removed the 7 patients who died from unknown causes and 120 patients who were missing first-year visit constancy or baseline CD4 count/plasma viral load, 1974 patients were included in the Cox proportional hazards model assessing factors associated with increased risk of overall (Table 4) and AIDS-related and non–AIDS-related (Table 5) mortality. Higher visit constancy during the first year of follow-up (per 20% increase; adjusted hazard ratio [HR] 0.87, 95% CI 0.79–0.96), being MSM (adjusted HR 0.50, 95% CI 0.29–0.87) and having a higher baseline CD4 count (adjusted HR 0.85, 95% CI 0.80–0.91) were associated with a lower risk of overall mortality, and higher visit constancy during the first year of follow-up was associated with a lower risk of AIDS-related death (per 20% increase; adjusted HR 0.79, 95% CI 0.64–0.97).

View this table:
[Table 4:](http://www.cmajopen.ca/content/7/2/E236/T4)

Table 4: 
Cox proportional hazards model for overall deaths among clinic patients, 2004–2015* (*n* = 1974, of whom 190 died and had documented cause of death)

View this table:
[Table 5:](http://www.cmajopen.ca/content/7/2/E236/T5)

Table 5: 
Cox proportional hazards model for AIDS-related and non–AIDS-related deaths among clinic patients, 2004–2015* (*n* = 1974)

## Interpretation

We describe retention in care and mortality trends among people with HIV infection who accessed a comprehensive care HIV clinic in an urban setting from 2004 to 2015. The overall probability of survival was significantly increased for patients with higher retention in care. Although the primary underlying cause of death was AIDS-related (28%), half of these deaths occurred in the earlier years of the study (2004–2007). Between 2008 and 2015, 54% of all deaths were non–AIDS-related, with non-AIDS malignant disease and substance use accounting for 33% of all deaths.

These finding are consistent with those of other studies conducted in similar settings25,26 and with population-based studies conducted in BC.27,28 The proportion of HIV-related deaths varied between studies, as the studies assessed different periods of time or people with different demographic characteristics. Trickey and colleagues29 reported that, among people with HIV infection who survived for 10 years or more after starting ART, death was more likely to result from non–AIDS-related nonliver cancer than from AIDS-related causes (25% v. 16%). In our study, non–AIDS-related nonliver cancers and substance use were the most commonly reported non–AIDS-related causes of death and the most commonly reported causes of death after 2007. Together, these findings suggest that, among HIV-infected people receiving long-term ART, non–AIDS-related nonliver cancers are a major cause of mortality. The decline in AIDS-related mortality and AIDS diagnoses over the study period in the current study corresponds to local efforts in the expansion and early uptake of ART,8,30,31 and the increase in deaths due to non-AIDS malignant disease is consistent with previous studies.27,28,32,33

Non–AIDS-related comorbidities including substance use and psychiatric conditions may reduce retention in care among people with HIV infection.34 In addition, providing support services for people with HIV infection improves ART adherence.35 Our findings suggest that access to a comprehensive care clinic that provides interdisciplinary services for HIV-infected people can lead to a substantially higher probability of overall survival (and a reduction in AIDS-related mortality). Our results align with those of Sherer and colleagues36 showing that complementing HIV-infection–related services with services not related to HIV infection is associated with better retention and better outcomes in this population.

### Limitations

This study has several limitations. First, it was an observational study, and causation cannot be inferred. Second, the retrospective nature of the study led to incomplete data capture and constraint on the variables available for analysis. Several variables (e.g., chronic hepatitis B [*n* = 10 among deceased patients]) were excluded from the survival analysis owing to low counts or collinearity with other variables (e.g., hepatitis C virus infection and injection drug use, and year of clinic enrolment and visit constancy exhibited collinearity in our study population37). Also, a single research assistant was responsible for clinical data collection. To minimize the possibility of measurement error, data extraction for medical conditions was standardized based on given operational definitions. Any discharge records or autopsy reports included were transcribed directly from the medical records without individual interpretation from the research assistant. In addition, the research assistant received training in data extraction and had various practice sessions to ensure that the data collected for the reviews were as accurately transcribed as possible. Third, it should be noted that the Cox proportional hazards model for AIDS-related and non–AIDS-related mortality excluded deceased patients whose cause of death was unknown and those with less than 1 year of follow-up in the clinic, which reduced the sample size from 271 to 190. The 81 patients (29.9%) excluded likely were not attending the clinic or were presenting late for care. Fourth, the relation between retention in care and death is affected by the increments (continuous) or threshold (dichotomized) used to delineate visit constancy. The threshold of 75% that we used was significantly associated with decreased overall, AIDS-related and non–AIDS-related mortality in multivariate modelling (data not shown). This value, however, has not been validated or used in the literature to indicate retention in care. To keep with existing literature, we expressed visit constancy as a continuous variable, in 20% increments; narrowing the increments further would have resulted in a slight weakening of the inverse relation between visit constancy and death, whereas wider increments would have yielded a stronger relation. We observed that the greatest reduction in all types of mortality occurred when patients attended at least 50% of clinic visits in their first year of follow-up (Figure 2). Further exploration to determine the threshold that best indicates retention in care is needed, particularly in relation to marginalized and stigmatized populations, who experience high barriers to clinic attendance.38 Fifth, the CoDe protocol has been previously evaluated, with most categories showing high initial agreement12; however, several categories that we used in the present study, including non–AIDS-related infection, chronic obstructive lung disease, psychiatric disease, and heart or vascular disease, yielded 43%–56% agreement. The low agreement in these categories may have been due to small samples (e.g., 2 cases of chronic obstructive lung disease, with disagreement on 1 case). Our determinations of cause of death are bolstered by the high interrater reliability (89%) between physicians. Finally, although the study highlights trends in mortality among people with HIV infection and the impact of retention in comprehensive HIV infection care, caution should be used in generalizing the findings to other vulnerable populations. However, similar results may be observed in other clinics where ART can be accessed free of charge and comprehensive, interdisciplinary care for marginalized and vulnerable people with HIV infection is provided.

### Conclusion

Higher retention in care was associated with a lower risk of death in our study. On the other hand, non–AIDS-related causes accounted for a higher number of deaths during the later years of the study. Subsequent studies should examine how various elements of comprehensive care models benefit people infected with HIV, how these models can be implemented on a larger scale and how they compare with models of nonintegrated HIV infection care.

## Acknowledgements

The authors acknowledge the technical support provided by staff at the BC Centre for Excellence in HIV/AIDS, particularly Nada Gataric, Raja Dutta and Angie Semple, for their technical support. They also thank Elizabeth Quayle for proofreading the manuscript and all the patients, physicians, allied health care professionals and administrative staff at the John Ruedy Immunodeficiency Clinic.

## Footnotes

*   **Competing interests:** Cole Stanley reports personal fees from ViiV Healthcare, Merck and Gilead Sciences outside the submitted work. Viviane Lima reports grants from the Canadian Institutes of Health Research (CIHR) and research fellowships from the CIHR and the Michael Smith Foundation for Health Research outside the submitted work. Julio Montaner reports grants paid to his institution from the Public Health Agency of Canada, the BC Ministry of Health, Gilead Sciences, Merck and ViiV Healthcare, a Knowledge Translation Award paid to his institution from the CIHR and a grant from the US National Institute on Drug Abuse, National Institutes of Health, outside the submitted work. Silvia Guillemi reports participation in advisory board meetings and meeting honoraria from Gilead Sciences, and personal fees from Gilead Sciences and ViiV Healthcare outside the submitted work. No other competing interests were declared.

*   This article has been peer reviewed.

*   **Contributors:** Rolando Barrios, Silvia Guillemi, Viviane Lima and Julio Montaner conceived and designed the study. Ana Ulloa collected the data. Benita Yip, Viviane Lima and Wendy Zhang analyzed the data. Ana Ulloa interpreted the data and drafted the manuscript. Cathy Puskas revised the analyses and manuscript according to reviewer feedback. All of the authors critically revised the manuscript for important intellectual content, approved the version to be published and agreed to be accountable for all aspects of the work.

*   **Supplemental information:** For reviewer comments and the original submission of this manuscript, please see [www.cmajopen.ca/content/7/2/E236/suppl/DC1](http://www.cmajopen.ca/content/7/2/E236/suppl/DC1).

## References

1.  Samji H, Cescon A, Hogg RS, et al. (2013) Closing the gap: increases in life expectancy among treated HIV-positive individuals in the United States and Canada. PLoS One 8:e81355.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1371/journal.pone.0081355&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=24367482&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 

2.  Irvine MK, Chamberlin SA, Robbins RS, et al. (2015) Improvements in HIV care engagement and viral load suppression following enrollment in a comprehensive HIV care coordination program. [published erratum in *Clin Infect Dis* 2015;60:1879]. Clin Infect Dis 60:298–310.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1093/cid/ciu783&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=25301208&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 

3.  Backus L, Czarnogorski M, Yip G, et al. (2015) HIV care continuum applied to the US Department of Veterans Affairs: HIV virologic outcomes in an integrated health care system. J Acquir Immune Defic Syndr 69:474–80.
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=25835603&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 

4.  Simeone C, Shapiro B, Lum PJ (2017) Integrated HIV care is associated with improved engagement in treatment in an urban methadone clinic. Addict Sci Clin Pract 12:19.
    
    

5.  Soto TA, Bell J, Pillen MB, HIV/AIDS Treatment Adherence, Health Outcomes and Cost Study Group (2004) Literature on integrated HIV care: a review. AIDS Care 16(Suppl 1):43–55.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1080/09540120412331315295&link_type=DOI) 

6.  Zaller N, Gillani FS, Rich JD (2007) A model of integrated primary care for HIV-positive patients with underlying substance use and mental illness. AIDS Care 19:1128–33.
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=18058396&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 

7.  (2015) HIV and AIDS in Canada: surveillance report to December 31, 2014 (Public Health Agency of Canada, Ottawa).
    
    

8.  Olding M, Enns B, Panagiotoglou D, et al. (2017) A historical review of HIV prevention and care initiatives in British Columbia, Canada: 1996–2015. J Int AIDS Soc 20:21941.
    
    

9.  Patterson S, Cescon A, Samji H, et al. (2015) Cohort profile: HAART Observational Medical Evaluation and Research (HOMER) cohort. Int J Epidemiol 44:58–67.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1093/ije/dyu046&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=24639444&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 

10. Hogg RS, Yip B, Chan KJ, et al. (2001) Rates of disease progression by baseline CD4 cell count and viral load after initiating triple-drug therapy. JAMA 286:2568–77.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1001/jama.286.20.2568&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=11722271&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000172385200031&link_type=ISI) 

11. CoDe Working Group (2013) Protocol: Coding Causes of Death in HIV (CoDe). Version 2.3 (Centre of Excellence for Health, Immunity and Infections, Copenhagen) Available: [https://chip.dk/Portals/0/files/Code%20Protocol%202.3.pdf](https://chip.dk/Portals/0/files/Code%20Protocol%202.3.pdf). accessed 2016 Apr 4.
    
    

12. Kowalska JD, Friis-Møller N, Kirk O, et al. (2011) The Coding Causes of Death in HIV (CoDe) Project: initial results and evaluation of methodology. Epidemiology 22:516–23.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1097/EDE.0b013e31821b5332&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=21522013&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000291252100013&link_type=ISI) 

13. Mugavero MJ, Davila JA, Nevin CR, et al. (2010) From access to engagement: measuring retention in outpatient HIV clinical care. AIDS Patient Care STDS 24:607–13.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1089/apc.2010.0086&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=20858055&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000282682300001&link_type=ISI) 

14. Yehia BR, Fleishman JA, Metlay JP, et al. (2012) Comparing different measures of retention in outpatient HIV care. AIDS 26:1131–9.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1097/QAD.0b013e3283528afa&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=22382143&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000304364900009&link_type=ISI) 

15. Therapeutic Guidelines Committee Therapeutic guidelines (BC Centre for Excellence in HIV/AIDS, Vancouver) Available: [www.cfenet.ubc.ca/therapeutic-guidelines](http://www.cfenet.ubc.ca/therapeutic-guidelines). accessed 2018 Jan 15.
    
    

16. Grossberg R, Zhang Y, Gross R (2004) A time-to-prescription-refill measure of antiretroviral adherence predicted changes in viral load in HIV. J Clin Epidemiol 57:1107–10.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1016/j.jclinepi.2004.04.002&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=15528063&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000225276100014&link_type=ISI) 

17. Steiner JF, Prochazka AV (1997) The assessment of refill compliance using pharmacy records: methods, validity, and applications. J Clin Epidemiol 50:105–16.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1016/S0895-4356(96)00268-5&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=9048695&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=A1997WF66800012&link_type=ISI) 

18. Low-Beer S, Yip B, O’Shaughnessy MV, et al. (2000) Adherence to triple therapy and viral load response. J Acquir Immune Defic Syndr 23:360–1.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1097/00126334-200004010-00016&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=10836763&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000087008400014&link_type=ISI) 

19. Genberg BL, Wilson IB, Bangsberg DR, et al. (2012) Patterns of antiretroviral therapy adherence and impact on HIV RNA among patients in North America. AIDS 26:1415–23.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1097/QAD.0b013e328354bed6&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=22767342&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000306130700012&link_type=ISI) 

20. Lima VD, Gill VS, Yip B, et al. (2008) Increased resilience to the development of drug resistance with modern boosted protease inhibitor-based highly active antiretroviral therapy. J Infect Dis 198:51–8.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1086/588675&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=18498238&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000256632800011&link_type=ISI) 

21. McCullagh P, Nelder J (1989) Generalized linear models, no 37 of *Monographs on statistics and applied probability* series (Chapman and Hall/CRC, Boca Raton (FL)), 2nd ed. pp 193–236.
    
    

22. Viera AJ, Garrett JM (2005) Understanding interobserver agreement: the kappa statistic. Fam Med 37:360–3.
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=15883903&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000229008800015&link_type=ISI) 

23. Cox DR (1972) Regression models and life-tables. J R Stat Soc B 34:187–220.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1111/j.2517-6161.1972.tb00899.x&link_type=DOI) 

24. 1.  Parzen E, 
    2.  Tanabe K, 
    3.  Kitagawa G
    
    Akaike H (1973) in Selected papers of Hirotugu Akaike. Springer series in statistics (Perspectives in statistics), Information theory and an extension of the maximum likelihood principle, eds Parzen E, Tanabe K, Kitagawa G (Springer, New York), pp 267–81.
    
    

25. Deans GD, Raffa JD, Lai C, et al. (2013) Mortality in a large community-based cohort of inner-city residents in Vancouver, Canada. CMAJ Open 1:E68–76.
    
    

26. DeLorenze GN, Weisner C, Tsai AL, et al. (2011) Excess mortality among HIV-infected patients diagnosed with substance use dependence or abuse receiving care in a fully integrated medical care program. Alcohol Clin Exp Res 35:203–10.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1111/j.1530-0277.2010.01335.x&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=21058961&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 

27. Cheung CC, Ding E, Sereda P, et al. (2016) Reductions in all-cause and cause-specific mortality among HIV-infected individuals receiving antiretroviral therapy in British Columbia, Canada: 2001–2012. HIV Med 17:694–701.
    
    

28. Eyawo O, Franco-Villalobos C, Hull MW, et al. (2017) Changes in mortality rates and causes of death in a population-based cohort of persons living with and without HIV from 1996 to 2012. BMC Infect Dis 17:174.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1186/s12879-017-2254-7&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=28241797&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 

29. Trickey A, May MT, Vehreschild J, et al. (2016) Cause-specific mortality in HIV-positive patients who survived ten years after starting antiretroviral therapy. PLoS One 11:e0160460.
    
    

30. Montaner JS, Lima VD, Barrios R, et al. (2010) Association of highly active antiretroviral therapy coverage, population viral load, and yearly new HIV diagnoses in British Columbia, Canada: a population-based study. Lancet 376:532–9.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1016/S0140-6736(10)60936-1&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=20638713&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000281266800032&link_type=ISI) 

31. Montaner JS, Lima VD, Harrigan PR, et al. (2014) Expansion of HAART coverage is associated with sustained decreases in HIV/AIDS morbidity, mortality and HIV transmission: the “HIV treatment as prevention” experience in a Canadian setting. PLoS One 9:e87872.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1371/journal.pone.0087872&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=24533061&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 

32. Ingle SM, May MT, Gill MJ, et al. (2014) Impact of risk factors for specific causes of death in the first and subsequent years of antiretroviral therapy among HIV-infected patients. Clin Infect Dis 59:287–97.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1093/cid/ciu261&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=24771333&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 

33. Smith CJ, Ryom L, Weber R, et al., D:A:D Study Group (2014) Trends in underlying causes of death in people with HIV from 1999 to 2011 (D:A:D): a multicohort collaboration. Lancet 384:241–8.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1016/S0140-6736(14)60604-8&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=25042234&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000339384100025&link_type=ISI) 

34. Altice FL, Kamarulzaman A, Soriano VV, et al. (2010) Treatment of medical, psychiatric, and substance-use comorbidities in people infected with HIV who use drugs. Lancet 376:367–87.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1016/S0140-6736(10)60829-X&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=20650518&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 
    
    [Web of Science](http://www.cmajopen.ca/lookup/external-ref?access_num=000280717900035&link_type=ISI) 

35. Higa DH, Marks G, Crepaz N, et al. (2012) Interventions to improve retention in HIV primary care: a systematic review of U.S. studies. Curr HIV/AIDS Rep 9:313–25.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1007/s11904-012-0136-6&link_type=DOI) 
    
    [PubMed](http://www.cmajopen.ca/lookup/external-ref?access_num=22996171&link_type=MED&atom=%2Fcmajo%2F7%2F2%2FE236.atom) 

36. Sherer R, Stieglitz K, Narra J, et al. (2002) HIV multidisciplinary teams work: support services improve access to and retention in HIV primary care. AIDS Care 14(Suppl 1):31–44.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1080/09540120220149975&link_type=DOI) 

37. Patrick DM, Tyndall MW, Cornelisse PG, et al. (2001) Incidence of hepatitis C virus infection among injection drug users during an outbreak of HIV infection. CMAJ 165:889–95.
    
    [Abstract/FREE Full Text](http://www.cmajopen.ca/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NDoiY21haiI7czo1OiJyZXNpZCI7czo5OiIxNjUvNy84ODkiO3M6NDoiYXRvbSI7czoyMDoiL2NtYWpvLzcvMi9FMjM2LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

38. Mahajan AP, Sayles JN, Patel VA, et al. (2008) Stigma in the HIV/AIDS epidemic: a review of the literature and recommendations for the way forward. AIDS 22:S57–65.
    
    [CrossRef](http://www.cmajopen.ca/lookup/external-ref?access_num=10.1097/01aids.0000341777.78876.40&link_type=DOI) 

*   Copyright 2019, Joule Inc. or its licensors