From the field in East Sumba. Medical care, clean water, prevention and community health where access remains difficult.
Sumba Sepsis Study 2026: Interactive Results
The Sumba Sepsis Study 2026 turned months of fieldwork into evidence that can now be explored. See where infection prevention fails, where time is lost before care, what healthcare workers report, and which actions the findings now support.
Explore the Sumba Sepsis Study results
1,001 households, 162 healthcare-worker submissions and one question: where is effective care being lost?
The Sumba Sepsis Study 2026 follows the steps between preventing infection at home and reaching effective care in East Sumba. Start with the six main findings, then compare household responses across eight localities. The healthcare-worker survey gives a separate view of what happens when people seek care.
Read each percentage with its denominator, because not everyone answered every question. For example, the November analysis of overlapping prevention conditions uses 939 of the 1,001 household records. Reports of severe illness describe what households recalled; they do not establish how many people had sepsis.
The Sumba Sepsis Study film shows the people and roads behind the data.
Household fieldwork: 2 June–17 August 2026. Nine household-size answers were missing. Residents were represented through household reports, not individual examinations.
A person must remain visible behind every record. These findings describe prevention, reported illness and access to care in the surveyed households. They do not estimate sepsis prevalence or prove that a programme improved survival.
Start with the essential findings
Six findings that call for practical action
The six headline findings are unchanged in report v1.1. Each keeps its own denominator and its source. Open a card to understand what the number measures.
Report v1.1, p. 3 · Household reports and staff submissions are separate sources.
No water and soap together at a handwashing place
950 of 998 household responses
About this result
This describes the reported absence of water and soap together at the household handwashing place. It does not mean that families never wash their hands, and the condition was not independently verified in every home.
HH-Q19 · Report p. 13
Staff reported patients arriving late in advanced condition
158 of 162 healthcare-worker submissions said this occurs at least sometimes
About this result
This is a healthcare-worker questionnaire result. It does not mean that 97.5% of patients arrive late and is not a patient-level rate.
HW-Q21 · Report p. 20
No previous sepsis or severe-infection training
144 of 162 healthcare-worker submissions
About this result
The finding documents reported prior topic-specific training. It is not a direct assessment of individual clinical competence.
Healthcare-worker survey · Report p. 18
Injectable antibiotics were not always available in primary care
70 of 105 primary-care submissions reported sometimes, rarely or never
About this result
These are staff reports, not audited stock records or percentages of health facilities. They identify concern about reliable treatment at first contact.
Healthcare-worker survey · Report p. 19
Households faced prevention and access barriers at the same time
Every household reporting at least one hour to formal care also lacked water and soap together
About this result
The same 286 household records carried both findings. This is a descriptive overlap, not evidence that one barrier caused the other.
HH-Q20 + HH-Q19 · Report p. 14
No toilet or open defecation reported
401 of 996 household responses
About this result
Simple pit latrines were not classified as open defecation. Construction quality was not assessed.
HH-Q18 · Report p. 13
Household evidence · eight indicators
What do families face, and where?
Choose a condition, then a locality. Compare the recorded responses, with the number answering each question beside the percentage. The locality selection applies only to this household explorer.
No water and soap together
95.2% 950 of 998 household responses
3 missing answers out of 1,001 households.
Reported absence of water and soap together at a handwashing place. This does not mean that families never wash their hands. Conditions were not independently verified in every home.
No toilet / open defecation
40.3% 401 of 996 household responses
5 missing answers out of 1,001 households.
Simple pit latrines are not counted as open defecation. Construction quality and safe management were not assessed.
At least one hour to formal care
28.7% 286 of 997 household responses
4 missing answers out of 1,001 households.
Usual travel to the nearest formal facility, not a verified emergency journey or time to definitive treatment. Six seasonal answers remain in the 997 responses, as a separate category; they are not counted as journeys under one hour.
Smoking inside the home
82.5% 814 of 987 household responses
14 missing answers out of 1,001 households.
Someone usually smoking indoors was reported by these households. This is a household condition, not the proportion of people who smoke or a measured exposure level.
Firewood mentioned for cooking
98.0% 974 of 994 household responses
7 missing answers out of 1,001 households.
Includes firewood mentioned alongside another fuel. There are 67 multiple-fuel answers. This does not establish exclusive firewood use, indoor cooking or measured air pollution.
Open well, surface water or river
25.7% 257 of 999 household responses
2 missing answers out of 1,001 households.
Main drinking-water source as reported. Source type is not a microbiological test and does not establish whether water is safe.
No water treatment reported alone
40.8% 406 of 995 household responses
6 missing answers out of 1,001 households.
Counts no treatment only when no treatment method was also selected. Another 49 responses combined no treatment with boiling or filtering; they stay separate. No treatment alone is not proof of unsafe water or of a need for treatment.
Severe or frightening illness reported
24.9% 249 of 1,001 household responses
0 missing answers out of 1,001 households.
Reported household illness in the previous six months, without clinical verification. This is not sepsis prevalence. A negative answer does not establish absence of illness.
These are sample descriptions, not village sepsis-risk rankings. Selection, recall and interviewing may affect differences. Mbinudita / Praipaha is counted once. Missing answers are not treated as negative. Locality breakdowns were checked against the preserved household workbook on 12 November 2026.
Explore the village map and more household questions in the dashboard ↗
New in report v1.1 · Appendix D
See the needs that overlap in a home
The November analyses revisit the same household records. Counting specified conditions helps identify what needs checking and improving. It does not predict who will develop sepsis.
Prevention conditions · Q16–Q19
Three or four counted conditions
32.8%
308 of 939 classifiable households
Water source, water treatment, sanitation and water with soap are considered together.
62 households remain unclassified. Missing or contradictory answers were not assigned zero.
See the counting rule and all groups
One point for each reported condition: open well/surface water or river/stream; no water treatment alone; no toilet/open defecation; no handwashing place with water and soap together.
| Conditions | Households | Share |
|---|---|---|
| 0–1 | 298 | 31.7% |
| 2 | 333 | 35.5% |
| 3–4 | 308 | 32.8% |
Exact counts from 0 to 4: 17, 281, 333, 258 and 50. Equal points are a counting rule, not equal biological effects. Water source and treatment are related measures. No treatment alone does not prove unsafe water; zero or one condition does not prove protection.
Access obstacles · Q20–Q21
Long travel and transport considered separately
301 households
301 of 991 (30.4%) had one counted obstacle; nine (0.9%) had both.
One obstacle for a journey of at least one hour, and one for no reliable transport. Distance is not added again for the same journey.
10 households remain unclassified, including six with strongly seasonal travel times.
See all access groups
| Obstacles | Households | Share of 991 |
|---|---|---|
| 0 | 681 | 68.7% |
| 1 | 301 | 30.4% |
| 2 | 9 | 0.9% |
Zero means neither of these two recorded barriers, not easy access in every circumstance. This count is separate from prevention and is not a clinical triage score.
Smoke in the home: what is established, and what still needs clarification?
Indoor smoking: 814/987 (82.5%). Firewood mentioned: 974/994 (98.0%). These two results stand independently. They do not measure inhaled smoke, ventilation or respiratory illness.
The kitchen-location export combines “inside the main house / less than 5 m” in one label, selected in 593/996 responses (59.5%). The questionnaire separates those locations, so the data cannot establish that all 593 kitchens were indoors.
501 records contain all three positive entries. One exclusion rule yields 501/979 (51.2%), as displayed by the reviewed dashboard; another yields 501/969 (51.7%). The difference concerns ten non-firewood records with follow-up answers. No single combined percentage is treated as a settled headline here.
Compare indoor smoking by locality · Read the sensitivity check · p. 36 ↗
From the home to follow-up
Follow the journey to effective care
A family needs practical prevention, recognition of deterioration, a way to reach help and a service able to respond. The surveys describe different people; these stages are not linked patient histories.
Make prevention possible at home
Water, soap, sanitation and the conditions in which families cook matter before anyone becomes seriously ill. Health information needs to be accompanied by the means to act on it.
Check which conditions coexist in each home. The new prevention count describes needs; it does not predict sepsis.
Explore the November analyses →
Report pp. 13 and 35–36
Notice when a familiar illness becomes dangerous
249 of 1,001 households reported severe or frightening illness in the preceding six months. These reports were not clinically confirmed as sepsis.
Field accounts raised concern that recurrent illness can become familiar. The study does not measure how common under-recognition was.
Read the human context of Q23 →
Report pp. 15 and 26
Trust does not remove the barriers to help
993 of 995 households reported full or some confidence in formal care. Separately, 452 of 999 responses described seeking outside help on the same day after illness became serious.
These different questions are not a linked patient timeline. Cost, distance and transport can stand between wanting care and obtaining it.
Report p. 17
The first journey can already be long
286 of 997 households reported at least one hour to the nearest formal health facility; 71 reported more than two hours. All 286 also lacked a reported handwashing place with water and soap together.
This is a descriptive overlap in the same household records, not evidence that one barrier caused the other. Seasonal journeys remain visible in the analysis.
Compare travel times by locality →
Report p. 14
Make the whole treatment pathway reliable
70 of 105 primary-care submissions reported injectable antibiotics only sometimes, rarely or never available. Oxygen, IV fluids and monitoring were more commonly reported.
Assessment, medicines, monitoring, reassessment and escalation need to work together. Staff questionnaires identify concerns; stock and equipment checks are still needed.
Report pp. 18–19
Follow the transfer through to arrival
Among 137 staff submissions from respondents who refer severe patients, 83 selected decision-making delay, 71 distance, 56 receiving-facility overcrowding and 48 cost. Several answers could be selected.
The question does not identify whose decision was delayed. A referral recommendation is not a verified completed transfer.
See all reported referral barriers →
Report p. 20
Keep the person visible after treatment
63 of 246 selected episodes with an outcome answer involved incomplete recovery or lasting weakness/disability: 54 were still ill or not fully recovered, and nine involved permanent weakness or disability.
Three of the 249 eligible households had no outcome answer. These reported outcomes do not establish post-sepsis syndrome. Prospective follow-up is needed.
Connect the finding to practical action →
Report p. 16
Listening to household experience
A familiar symptom can still need attention
Reported severe or frightening illness within six months.
Field accounts describe families for whom recurrent cough, diarrhoea, weakness or skin infection had become familiar. Serious illness could mean extreme deterioration or hospitalisation.
Those accounts raise a concern about under-recognition. The survey does not measure how often that happened. The 752 households that did not report a severe episode cannot be assumed free of illness.
HH-Q23 · Report pp. 15 and 26. Reported experience is distinct from a clinical diagnosis.
Keep the sources distinct
Three views of the care system
Household interviews, staff questionnaires and administrative returns answer different questions. A village filter on household data does not filter the staff findings.
Reported needs and decisions at home
Timing answers were not consistently linked to the selected illness episode. Medicine purchases do not establish antibiotic misuse: drug class, dose and indication were not reliably established. These results describe the sample, not all of East Sumba.
Reported first-contact readiness
There were 162 submissions, including 105 from primary care. 144/162 reported no prior sepsis or severe-infection training, while 158/162 reported late presentation in advanced condition at least sometimes.
Other reported resources and denominator limits
IV fluids: 101/105; basic monitoring: 95/105; basic haematology: 32/105. Several staff may describe the same establishment. One possible repeated name/facility pair remains unresolved; no submission was silently removed.
Staff percentages are proportions of questionnaire submissions, not patients or independently audited facilities. No prior topic-specific training is not a direct test of clinical competence.
What remains uncounted
The supplied infection return listed 28 facilities. Zeros accounted for 232/308 possible infection entries in 2025 and 228/308 in 2026. Seven diagnostic categories were zero throughout both columns.
Zero is not the same as no disease. Definitions, reporting completeness, population denominators and the 2026 cut-off were not established. These returns cannot estimate district sepsis burden or infection-attributable mortality. The capacity form also carries an unresolved date entry.
137 staff submissions from those referring severe patients
Where referral becomes difficult
Several barriers could be selected. These percentages are mentions, not the share of delays caused by each factor.
Report p. 20. The question does not establish whose decision was delayed. These barriers overlap and percentages need not add to 100%.
From evidence to action
Six priorities emerging from the study
These are proposed responses derived from the evidence. The study did not test their effectiveness. Each priority therefore includes a practical action and a way future progress could be measured.
1. Material prevention
Evidence behind this priority: 950 of 998 household responses reported no water and soap together at a handwashing place; 401 of 996 reported no toilet or open defecation. Report p. 13.
Verify and improve access to water and soap, safe sanitation and household water practices in priority settlements.
Measure: functioning handwashing place, water and soap availability, and sanitation use.
2. Recognition and first contact
Evidence behind this priority: 144 of 162 healthcare-worker submissions reported no previous sepsis or severe-infection training. Field accounts also raised concern about illness recognition, without quantifying under-recognition. Report pp. 18 and 26.
Build supervised, locally understandable danger-sign assessment and practical clinical skills.
Measure: demonstrated skills, appropriate escalation and missed deterioration on review.
3. Transport and referral
Evidence behind this priority: 286 of 997 households reported at least one hour to formal care. Among 137 staff submissions from those referring severe patients, 83 selected decision-making delay and 71 distance. Report pp. 14 and 20.
Clarify contact and acceptance arrangements, available drivers or vehicles, and support for families facing transport costs.
Measure: decision, departure and arrival times, completed transfer and recorded outcome.
4. Treatment readiness
Evidence behind this priority: 70 of 105 primary-care submissions described injectable antibiotics as available sometimes, rarely or never. These are staff reports, not independently audited stock records. Report p. 19.
Verify level-appropriate medicines, oxygen delivery, monitoring and trained shift coverage.
Measure: stock-out days, functional equipment, replenishment and staff coverage.
5. Diagnosis and safe medicine use
Evidence behind this priority: Blood-culture access was reported in 13 of 105 primary-care submissions, and basic haematology access in 32 of 105. These are reported resources, not verified availability for a specific patient. Report p. 19.
Clarify accessible tests, sample transport and clinician-led prescribing review.
Measure: usable results returned, treatment review and sample-transfer failures.
6. Count and follow severe illness
Evidence behind this priority: Referral counts were provided for only 2 of 24 Puskesmas in the supplied administrative workbook. Undefined zeros, reporting gaps and missing denominators prevent reliable burden or outcome estimates. Report p. 22.
Use defined categories and reporting periods, and distinguish zero, missing and not reported.
Measure: reporting completeness and reconciled severe episodes, transfers and outcomes.
Scientific safeguards
Read the results with their limits
One study, complementary ways to read it
Read the report. Explore the household data.
The report brings together methods, household results, staff responses, administrative findings and interpretation. The dashboard offers a village map, household comparisons and data-quality notes. Its kitchen-location label and combined-smoke denominator need the clarification in Appendix D.
Report presentation · Meet the people behind the study in the film · Methods and limitations
What changed in this page on 12 November 2026?
Updated all report links to v1.1; retained the six original headline findings; brought eight household indicators into one locality explorer; added the prevention and access counts from Appendix D; clarified the unsettled combined-smoke analysis; added direct links between findings, care stages, sources and actions.
Household data period: 2 June–17 August 2026. New descriptive analyses do not represent new interviews, a trend, a validated risk model or a measured intervention effect. English report page numbers are used throughout this page.
Report v1.1 · Updated 12 November 2026 · First published 13 September 2026.
Interface revision 5.0 · Epidemiological interpretation remains preliminary; no independent peer review is claimed.
English 
Français 
Deutsch (Schweiz) 
Italiano 