A comparison of the logistic risk function and the proportional hazards model in prospective epidemiologic studies

doi:10.1016/0021-9681(83)90165-0

Journal of Chronic Diseases

Volume 36, Issue 10, 1983, Pages 715-723

Journal of Chronic D...

https://doi.org/10.1016/0021-9681(83)90165-0 Get rights and content

Abstract

The logistic regression and proportional hazards models are each currently being used in the analysis of prospective epidemiologic studies examining risk factors in chronic disease applications. The advantages and disadvantages of each are yet to be fully described. However, a theoretical relationship between the two models has been documented.

In this paper the conditions under which results from the two models approximate one another are described. It is shown that where the follow-up period is short and the disease is generally rare, the regression coefficients of the logistic model approximate those of the proportional hazards model with a constant underlying hazard rate. Since under the same conditions the likelihood functions approximate one another, the regression coefficients have similar estimated standard errors. Further, estimation of relative risk with these models is constrasted.

These results are illustrated utilizing a previously published data set on metastatic cancer of the breast. With increasing follow-up time, the logistic regression coefficients become uncertain and less reliable.

References (17)

J Truett et al.
A multivariate analysis of the risk of coronary heart disease in Framingham
J Chron Dis
(1967)
T Gordon et al.
HDL cholesterol and CHD risk: The Framingham Study
Am J Med
(1977)
AR Dyer
An analysis of relationship of systolic blood pressures, serum cholesterol and smoking to 14-year mortality in the Chicago People's Gas Company Study
I. J Chron Dis
(1975)
J Cornfield
Joint dependence of risk of coronary heart disease on serum cholesterol and systolic blood pressure: A discriminant function analysis
U Goldbourt et al.
High density lipoprotein cholesterol and incidence of coronary heart disease: The Israeli Ischemic Heart Disease Study
Am J Epid
(1974)
D McGee et al.
The results of the Framingham study applied to four other U.S. based epidemiologic studies of cardiovascular disease
MH Myers et al.
A logistic-exponential model for use with response time data involving regression variables
Biometrics
(1973)
SJ Cutler et al.
Classification of patients with disseminated cancer of the breast
Cancer
(1969)

There are more references available in the full text version of this article.

Cited by (139)

Impact of beta-blockers on mortality and cardiovascular disease outcomes in patients with obstructive sleep apnoea: a population-based cohort study in target trial emulation framework
2023, The Lancet Regional Health - Europe
There is no real-world evidence regarding the association between beta-blocker use and mortality or cardiovascular outcomes in patients with obstructive sleep apnoea (OSA). We aimed to investigate the impact of beta-blocker use on all-cause mortality and cardiovascular diseases (CVDs) in patients with OSA.
We conducted a target trial emulation study of 37,581 patients with newly diagnosed OSA from 1st January 2000 to 30th November 2021 using the IMRD-UK database (formerly known as the THIN database). We compared the treatment strategies of initiating beta-blocker treatment within one year versus non-beta-blocker treatment through the method of clone-censor-weight. Covariates, including patients’ demographics, lifestyle, comorbidities, and recent medications, were measured and controlled. Patients were followed up for all-cause mortality or composite CVD outcomes (angina, myocardial infarction, stroke/transient ischaemic attack, heart failure, or atrial fibrillation). We estimated the five-year absolute risks, risk differences and risk ratio with 95% confidence intervals (CIs) with standardised, weighted pooled logistic regression, which is a discrete-time hazard model for survival analysis. Several sensitivity analyses were performed, including multiple imputation addressing the missing data.
The median follow-up time was 4.1 (interquartile range, 1.9–7.8) years. The five-year absolute risk of all-cause mortality and CVD outcomes were 4.9% (95% CI, 3.8–6.0) and 13.0% (95% CI, 11.4–15.0) among beta-blocker users, and 4.0% (95% CI, 3.8–4.2) and 9.4% (95% CI, 9.1–9.7) among non-beta-blocker users, respectively. The five-year absolute risk difference and risk ratio between the two groups for all-cause mortality and CVD outcomes were 0.9% (95% CI, −0.2 to 2.1) and 1.22 (95% CI, 0.96–1.54), and 3.5% (95% CI, 2.1–5.5) and 1.37 (95% CI, 1.22–1.62), respectively. Findings were consistent across the sensitivity analyses.
Beta-blocker treatment was associated with an increased risk of CVD and a trend for an increased risk of mortality among patients with OSA. Further studies are needed to confirm our findings.
Innovation and Technology Commission of the Hong Kong Special Administration Region Government.
Use of Time-to-Event Analysis to Develop On-Scene Return of Spontaneous Circulation Prediction for Out-of-Hospital Cardiac Arrest Patients
2022, Annals of Emergency Medicine
Citation Excerpt :
A systematic review found that studies of machine learning performance enhancement are mostly focused on discrimination, and calibration is not sufficiently reported.42 In previous studies, time-to-event analysis models yielded more precise estimates of association, likely leading to more precise prediction and greater statistical power than classification models would.43-45 We found that Utstein elements were effective in predicting time-to-on-scene return of spontaneous circulation.33
We aimed to train and validate the time to on-scene return of spontaneous circulation prediction models using time-to-event analysis among out-of-hospital cardiac arrest patients.
Using a Korean population-based out-of-hospital cardiac arrest registry, we selected a total of 105,215 adults with presumed cardiac etiologies between 2013 and 2018. Patients from 2013 to 2017 and from 2018 were analyzed for training and test, respectively. We developed 4 time-to-event analyzing models (Cox proportional hazard [Cox], random survival forest, extreme gradient boosting survival, and DeepHit) and 4 classification models (logistic regression, random forest, extreme gradient boosting, and feedforward neural network). Patient characteristics and Utstein elements collected at the scene were used as predictors. Discrimination and calibration were evaluated by Harrell’s C-index and integrated Brier score.
Among the 105,215 patients (mean age 70 years and 64% men), 86,314 and 18,901 patients belonged to the training and test sets, respectively. On-scene return of spontaneous circulation was achieved in 5,240 (6.1%) patients in the former set and 1,709 (9.0%) patients in the latter. The proportion of emergency medical services (EMS) management was higher and scene time interval longer in the latter. Median time from EMS scene arrival to on-scene return of spontaneous circulation was 8 minutes for both datasets. Classification models showed similar discrimination and poor calibration power compared to survival models; Cox showed high discrimination with the best calibration (C-index [95% confidence interval]: 0.873 [0.865 to 0.882]; integrated Brier score at 30 minutes: 0.060).
Incorporating time-to-event analysis could lead to improved performance in prediction models and contribute to personalized field EMS resuscitation decisions.
The association between self-reported mental health, medication record and suicide risk: A population wide study
2021, SSM - Population Health
Citation Excerpt :
Results are presented as odds ratio (OR) with 95% confidence interval (CI). As follow-up was short and the outcome rare, logistic regression was favoured over Cox Regression (Annesi et al., 1989; Callas et al., 1998; Green & Symons, 1983; Ingram & Kleinman, 1989). Approval for this study was obtained from the Office for Research Ethics Committees Northern Ireland (ORECNI), approval reference number 16/SC/0241.
Suicide mortality and mental ill health are increasing globally. Mental ill health can be measured in multiple ways. It is unclear which measure is most associated with suicide risk. This study explored the association between self-rated mental health and medication record and death by suicide. The 2011 Northern Ireland Census records of adults aged 18-74 years (n=1,098,967) were linked to a centralised database of dispensed prescription medication and death registrations until the end of 2015. Mental health status was ascertained through both a single-item self-reported question in the Census and receipt of psychotropic medication. Logistic regression models examined the association between indicators of mental ill health and likelihood of suicide mortality. Of the 1,098,967 cohort members, 857 died by suicide during the study period. Just over half of these deaths (n=429, 50.1%) occurred in individuals with neither indicator of mental ill health. Cohort members with both self-reported mental ill health and receipt of psychotropic medication had the highest risk of suicide (OR=6.13, 95%CI: 4.94–7.61), followed by those with psychotropic medication record only (OR=4.00, 95%CI: 3.28–4.88) and self-report only (OR=2.88, 95%CI: 2.16–3.84). Individuals who report mental ill health and have a history of psychotropic medication use are at a high risk of suicide mortality. However, neither measure is particularly sensitive, as both failed to signal over half of subsequent suicides. Some individuals who report poor mental health but are not in receipt of psychotropic medication are at increased risk of suicide, indicating possible unmet treatment need. The combination of the two indicators offers more precision for identifying those most at risk for targeted interventions.
Parental mental health and risk of poor mental health and death by suicide in offspring: a population-wide data-linkage study
2022, Epidemiology and Psychiatric Sciences
Effects of Warfarin on the Risks of Mortality, Acute Heart Failure, and Infection Resolution in Patients with Infective Endocarditis: A Target Trial Emulation
2024, SSRN
Tutorial on survival modeling with applications to omics data
2024, Bioinformatics

View all citing articles on Scopus

^∗: Present address: Rehov Hatsabar 6/13, Kiron, Israel.

View full text

Original articleA comparison of the logistic risk function and the proportional hazards model in prospective epidemiologic studies

Abstract

J Chron Dis

Am J Med

I. J Chron Dis

Joint dependence of risk of coronary heart disease on serum cholesterol and systolic blood pressure: A discriminant function analysis

High density lipoprotein cholesterol and incidence of coronary heart disease: The Israeli Ischemic Heart Disease Study

Am J Epid

The results of the Framingham study applied to four other U.S. based epidemiologic studies of cardiovascular disease

A logistic-exponential model for use with response time data involving regression variables

Biometrics

Classification of patients with disseminated cancer of the breast

Cancer

Original article
A comparison of the logistic risk function and the proportional hazards model in prospective epidemiologic studies