A retrospective study in Kano State, Nigeria, found no clear association between care setting and mortality among patients initially assessed as having mild diphtheria. That does not establish that home care and treatment-centre care are equivalent, or that home care reduced deaths. The comparison involved a structured, professionally triaged programme during an epidemic, not unsupervised care at home. Final PLOS Medicine study
For care-delivery teams, the useful question is how much a comparison can tell us when clinicians, capacity and household circumstances helped determine where patients were treated.
Who was compared
Médecins Sans Frontières introduced home-based care for mild disease in July 2023 as hospital capacity came under pressure. The study used registry records from four local-government areas and later interviews with patients or caregivers. The sampled patients had been notified between January 2023 and January 2024; interviews took place in September–October 2024.
The final analysis included 669 people: 362 managed through home-based care and 307 in diphtheria treatment centres. Patients were matched using age group, sex, vaccination status and area of residence, but treatment setting was not randomly assigned. Setting, eligibility and matching methods
“Mild” was a clinical classification at presentation, not a guarantee that everyone had the same underlying risk. The authors report that all recorded complications and comorbidities occurred among treatment-centre patients, suggesting that more complex presentations were more often admitted.
The main estimate was uncertain
For all-cause mortality, the primary adjusted odds ratio comparing home-based with treatment-centre care was 0.36, with a 95% confidence interval of 0.12–1.07. There were 24 deaths in the analysed cohort. Primary analysis
The point estimate favours home-based care, but the interval includes 1, the no-association value for an odds ratio. That does not prove no difference. It also does not convert the estimate into evidence that home care caused a reduction in mortality.
Some additional analyses produced more favourable estimates with intervals excluding 1. They did not all compare exactly the same patients or use the same adjustment. The paper explains that the conditional model used informative matched subclasses, whereas population-averaged analyses could include people in single-arm subclasses. Reading the most favourable result alone would hide that difference.
The approved protocol did not specify a formal non-inferiority hypothesis, and the final paper reports two-sided comparisons. It should therefore not be described as a trial proving that home care was no worse within an agreed acceptable margin.
Matching cannot remove an unrecorded severity difference
Matching aligns measured characteristics. It cannot guarantee balance on clinical details that were not captured adequately, or undo the clinician’s decision to admit someone whose presentation appeared more concerning.
Care setting was also related to calendar time because the home programme began partway through the epidemic. Adding time to an analysis did not resolve every uncertainty, especially with only 24 deaths. The authors explicitly identify residual confounding by severity and time as important limits. Discussion of confounding
Follow-up adds another selection issue. Only 491 of the 986 originally sampled people were interviewed. A further 178 substitute patients brought the analysed total to 669. Replacements were not specified in the approved protocol and sometimes came from a different matching stratum or care arm. The study describes these departures rather than presenting the final sample as complete follow-up of the original selection.
Household and longer-term outcomes need the same caution
The authors found no clear increase associated with home care in reported household transmission or long-term sequelae. These outcomes were uncommon or imprecise, and interviews typically occurred more than a year after notification. Household cases were based on reported compatible symptoms, rather than laboratory confirmation of every secondary case. Outcome definitions and limitations
Our Oxford hospital-at-home explainer concerns a different population and intervention. Neither result validates the other programme.
The Kano study contributes evidence about a specific response under capacity constraints. Decisions about triage, treatment, infection control or where an individual receives care require the applicable clinical and public-health arrangements; this historical comparison does not supply those instructions.