Covariance adjustment of rates based on the multiple logistic regression model

doi:10.1016/0021-9681(81)90040-0

Journal of Chronic Diseases

Volume 34, Issue 8, 1981, Pages 415-426

Journal of Chronic D...

https://doi.org/10.1016/0021-9681(81)90040-0 Get rights and content

Abstract

The multiple logistic regression method is gaining wide acceptance in the analysis of epidemiologic data where simultaneous adjustments for a number of confounding covariates is desired. However, the method is used primarily to determine the covariates-adjusted odds ratio, a statistic pertinent in epidemiologic research but not in the context of a comparative clinical trial. The main interest in a comparative clinical trial is to determine the covariates-adjusted difference in the ‘success’ (e.g. recovery) rate between therapeutic groups that is analogous to the statistical information obtained from analysis of covariance for quantitative response variables. By way of numeric examples, this paper provides a non-technical overviewing of the multiple logistic regression model, describes the meaning of the statistics obtained from multiple logistic regression analysis, and illustrates the computation of the covaries-adjusted ‘success’ rates for the respective therapies in a comparative clinical trial.

References (16)

K Yano et al.
Dietary intake and the risk of coronary heart disease in Japanese men living in Hawaii
Am J Clin Nutr
(1978)
OS Miettinen
Confounding and effect modification
Am J Epidemiol
(1974)
LG Dales et al.
An improper use of statistical significance testing in studying covariables
Int J Epidemiol
(1978)
P Armitage
Statistical Methods in Medical Research
(1971)
M Nerlove et al.
Univariate and multivariate log-linear and logistic models
J Neter et al.
Applied Linear Statistical Models
(1974)
JL Fleiss
Statistical Methods for Rates and Proportions
(1973)
N Mantel et al.
Tests for homogeneity of effects in an epidemiologic investigation
Am J Epidemiol
(1977)

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

Cited by (177)

Short-course versus long-course antibiotic treatment for uncomplicated vancomycin-resistant enterococcal bacteraemia: a retrospective multicentre cohort study
2023, Clinical Microbiology and Infection
Citation Excerpt :
Adjusted ORs with 95% CIs were derived from the logistic regressions. We also evaluated the marginal average effects [14–16]. To rule out the potential impact of differences in the survival times on effect estimates, we additionally provide person-time–adjusted effect estimates.
The optimal treatment duration for vancomycin-resistant enterococcal (VRE) bacteraemia is still a matter of debate. The aim of the present study was to compare short-course (≤9 days) and long-course (≥10 days) antibiotic treatments in hospitalized adult patients with uncomplicated VRE bacteraemia.
This retrospective study was conducted in four university hospitals in Germany. Adult patients with a positive blood culture for a VRE were screened from 1 January 2016 to 31 December 2018. Only patients who received a VRE-active antibiotic for at least 48 hours were included. The exclusion criteria were a survival of <10 days and a deep-seated source of infection requiring prolonged treatment. To compare the outcome of short-course therapy with that of long-course therapy, 30-day and 90-day overall mortality, relapse within 90 days, duration of hospitalization, and potential antibiotic-related adverse events were analysed by inverse probability of treatment weighting using the propensity score and by additional covariate adjustment.
Of the 363 patients screened, 219 (60.3%) patients were included in the final analysis. Among them, 48 (21.9%) patients had underlying haematological diseases. Seventy-eight (35.6%) patients received short-course treatment (median, 7 days; interquartile range, 5–8 days) and 141 (64.4%) patients received long-course treatment (median, 15 days; interquartile range, 12–23.5 days). Thirty-day mortality was similar in both groups (19.2% vs. 22.0%; adjusted OR, 1.15; p 0.773). Duration of hospitalization (in total and after onset of bacteraemia) was significantly shorter (p < 0.05) in the short-course treatment group, whereas other secondary outcome parameters did not differ between both groups.
Our study suggests that short-course treatment might not be associated with a worse outcome in patients with uncomplicated VRE bacteraemia.
Poor suitability for psychotherapy – a risk factor for treatment non-attendance?
2021, Journal of Affective Disorders
Patient suitability has been suggested to predict treatment non-attendance but information on its effect is limited.
To study the prediction of the Suitability for Psychotherapy Scale (SPS), on occurrence of treatment non-attendance.
Altogether 326 outpatients, with depressive or anxiety disorder, were randomized to short-term psychodynamic psychotherapy (SPP), long-term psychodynamic psychotherapy (LPP), and solution-focused therapy (SFT). SPS was based on seven components from three suitability domains: nature of problems, ego strength, and self-observing capacity. Treatment non-attendance was defined as refusal of engaging therapy and of premature termination. The Cox model and logistic regression were used.
Treatment non-attendance was significantly more common in LPP patients with poor SPS (RR = 2.76, 95% CI = 1.45-5.26). This was mainly due to poor flexibility of interaction, poor self-concept, and poor reflective ability. Premature termination in SFT showed a similar trend but due to other SPS components: absence of a circumscribed problem, poor modulation of affects, and poor response to trial interpretation. On the contrary, individuals with good values of SPS were more prone to premature termination in SPP.
The prediction of suitability on refusal could only be studied in the LPP group due to few refusals in the short-term therapy groups. The sample consisted of patients who participated in a trial. Thus the findings may not be directly generalized to unselected patients in the public mental health setting.
Poor suitability, apparently, predicts non-attendance in LPP and SFT, but not in SPP. More studies on large cohorts are needed.
Effectiveness of psychoanalysis and long-term psychodynamic psychotherapy on personality and social functioning 10 years after start of treatment
2019, Psychiatry Research
The evidence on potentially greater benefits of psychoanalysis (PA) vs. long-term psychodynamic psychotherapy (LPP) is scarce. This study compared the effectiveness of PA and LPP on personality and social functioning during a 10-year follow-up from the beginning of the treatments. The eligible patients, 41 self-selected for PA and 128 assigned to LPP, were 20–45 years of age and had anxiety or mood disorder. Outcomes were analyzed using ten standard measures of personality and social functioning, carried out 5–9 times during the follow-up. Different change patterns by time in PA and LPP emerged, suggesting less benefit of PA during the first years of follow-up and more benefit in most outcomes thereafter. Greater post-treatment improvement in PA than in LPP was seen up to 1–2 years after PA had ended in more mature defense style (DSQ), level of personality organization (LPO), more positive self-concept (SASB), more improved social adjustment (SAS-SR) and sense of coherence (SOC). However, at the 10-year follow-up the differences were non-significant. In conclusion, PA may give some additional benefits when long-term aims are linked to personality and social functioning. The relatively small differences and higher costs in comparison to LPP may restrict the feasibility of PA.
Childhood adversities as predictors of improvement in psychiatric symptoms and global functioning in solution-focused and short- and long-term psychodynamic psychotherapy during a 5-year follow-up
2018, Journal of Affective Disorders
Childhood adversities are frequent among adults who seek treatment for depression or anxiety. These disorders are commonly treated by psychotherapy. Yet it is not known if specific types or durations of psychotherapy are particularly suited for patients who have suffered various early adversities.
221 depressed and anxious adult outpatients from community, student, occupational, and private healthcare services filled the Childhood Family Atmosphere Questionnaire. They were randomly assigned to short- (solution-focused or psychodynamic) or long-term (psychodynamic) psychotherapy. Outcome was assessed via patient questionnaires and clinician interviews of psychiatric symptoms and global functioning during a 5-year follow-up. Linear regression analyses were conducted.
Less separations from caregivers expectedly predicted better outcomes in all therapies; unexpectedly, so did greater abuse. Family unhappiness and parental problems predicted faster or greater improvement when patients were assigned to a short- or long-term psychodynamic therapy model.
As patients with psychotic, substance abuse, and severe personality disorders were screened out, findings might not generalize to these patient groups.
Patients with certain childhood adversities appear to respond with faster or greater improvement when the psychotherapy model is explicitly focused on working through the potential connections between past and current problems. If confirmed by in-depth studies, the findings may help match psychotherapeutic models with given patient complaints, as well as fine-tune different psychosocial interventions to individual needs for optimizing treatment outcomes.
Dieting attempts modify the association between quality of diet and obesity
2017, Nutrition Research
Evidence on the nature of the relationship between obesity and the quality of diet remains controversial. Likewise, the possible effect of dieting attempts on this association is poorly understood. This study investigates the possible modifying effect of dieting attempts on the association between the quality of diet and obesity. The authors hypothesize that among dieters the association may be biased. The study was based on a Finnish cohort, including 5910 men and women aged 30 to 99 years, with information on diet and body mass index (BMI). Using data from a food frequency questionnaire (FFQ), an Alternate Healthy Eating Index (AHEI) applicable to the Finnish regimen was formed. Obesity was defined as a BMI ≥30 kg/m². Information on dieting attempts was collected using a questionnaire. The statistical analyses were based on linear and logistic regression. We found a positive association between the quality of the diet and obesity, the relative odds of obesity between the highest and lowest quintiles of AHEI being 1.48 (95% CI, 1.20-1.82) after adjustment for confounding factors. However, in the interaction analysis of dieting attempts and AHEI, no association was observed in non-dieters (OR, 0.92; 95% CI, 0.69-1.24) but among dieters a slightly elevated risk was found (OR, 1.40; 95% CI, 0.98-1.98). We found no association between a high quality diet and obesity among non-dieters, but a tendency for a positive association in dieters. Dieting thus seems to modify the association between diet and obesity, which should be further studied using a longitudinal design.
New Zealand Healthline call data used to measure the effect of travel time on the use of the emergency department
2017, Social Science and Medicine
Telephone triage is a health tool increasingly used to connect geographically distant populations. Such services are also utilised to address issues of Emergency Department (ED) overuse. New Zealand's tele-triage service, Healthline, has existed since 2001 but is yet to be the focus of analysis. This research sought to understand the role that travel time to ED had upon Healthline users' compliance with telephone advice. Additionally, the role of deprivation in Healthline (as a determinant of caller behaviour) was examined. Travel time to ED was found to influence the impact of Healthline advice upon callers but this was not confounded by deprivation. Those living closest to the ED were more likely to attend when advised to, and less likely to stay away if told to avoid the ED. Different time brackets showed stronger trends, suggesting that callers at varying distances from EDs may be more or less influenced by both travel time and Healthline advice.

View all citing articles on Scopus

View full text

Covariance adjustment of rates based on the multiple logistic regression model

Abstract

Am J Clin Nutr

Confounding and effect modification

Am J Epidemiol

An improper use of statistical significance testing in studying covariables

Int J Epidemiol

Statistical Methods in Medical Research

Univariate and multivariate log-linear and logistic models

Applied Linear Statistical Models

Statistical Methods for Rates and Proportions

Tests for homogeneity of effects in an epidemiologic investigation

Am J Epidemiol