Table 2.

Controlled before-after studies evaluating the effects of ACF on tuberculosis case notifications

	Country, population	Case-finding method	Diagnostic method	Co-interventions	Type of tuberculosis	Intervention group			Control group				Reported estimates
						Baseline CNR	Endline CNR	CNR ratio	Baseline CNR	Endline CNR	CNR ratio	Ratio of CNR ratios
Rendleman (1999)14	USA, people experiencing homelessness	Delivered alongside other services at shelters	TST for everyone; referral to clinician assessment with or without tests if TST positive	LTBI treatment	All types	227·4	96·9	0·43	3·94	4·67	1·19	0·36	None
de Vries et al (2007)15	Netherlands, people experiencing homelessness	Delivered alongside other services at shelters; mobile chest x-ray clinic	Chest x-ray regardless of symptoms; clinical assessment with or without culture if abnormal chest x-ray	None	All types	26·8	35·9	1·34	1·90	2·45	1·29	1·04	χ2 test for trend in 2002 to 2005 (ie, to show declining cases year on year after ACF introduced) in intervention population: p=0·03; no effect estimate comparing intervention to control population
Kan et al (2012)16	China, general population	Schoolchildren reporting symptoms of family members	Clinical review plus sputum smear if symptoms	Financial incentives and training to providers	Microbiologically confirmed	10·2	35·4	3·47	12·5	39·2	3·14	1·19	Case detection in counties receiving intervention increased by a factor of 3·5 compared with before intervention and by a factor of 3·1 compared with counties not receiving intervention (p=0·0001)*
Cegielski et al (2013)17	USA, general population	Door to door, community volunteers collecting and transporting sputum	TST for everyone; referral to clinician assessment with or without tests if TST positive	LTBI treatment	All types	47·6	0·0	0·00	7·29	4·84	0·66	0·00	Incidence declined from 15 cases (in 1985–1995) to zero cases (in 1996–2006) in the target neighborhoods, compared with 128 cases decreasing to 75 cases in the county overall (p=0·002)
Parija et al (2014)18	India, general population	Community mobilisation, mobile clinic, community health workers collecting and transporting sputum	Sputum smear if symptoms	None	Microbiologically confirmed	63·5	70·3	1·11	23·9	24·1	1·01	1·10	Number of smear-positive cases detected during the intervention period (April to June, 2012) increased by 11% relative to April to June, 2011, in intervention communities, compared with a 0·8% increase in non-intervention communities
Reddy et al (2015)19	India, indigenous populations plus informal urban	Door to door, community health workers collecting and transporting sputum	Sputum smear if symptoms	None	Microbiologically confirmed	60·5	65·8	1·09	50·7	46·4	0·91	1·19	Number of smear-positive cases detected increased by 8·8% relative to the pre-intervention period in intervention communities, compared with an 8·6% decrease in non-intervention communities
Sanaie et al (2016)20	Afghanistan, IDP camp	Door to door	Sputum smear if symptoms	Contact tracing, facility-based screening	Microbiologically confirmed	NA	NA	1·56†	NA	NA	0·75†	2·11	Comparison of trend in notifications over time in intervention area clinics and state; projecting the declining secular trend of notifications to 2012, only 59% of cases (2885 cases; 95% CI 2129–3640) notified during the intervention would have been notified without the intervention
Delva et al (2017)21	Haiti, IDP camp	Door to door, community health workers collecting and transporting sputum	Sputum smear if symptoms (Xpert at one of four sites)	Contact tracing, laboratory strengthening, facility-based screening	Microbiologically confirmed	33·5	53·5	1·59	30·9	34·8	1·13	1·42	Annual sputum smear-positive, bacteriologically positive notification rate in intervention population increased from 34 per 100 000 individuals to 54 per 100 000 (59% increase, 95% CI 4 to 143; p=0·03); in the control population, the notification rate was 31 per 100 000 before intervention and 35 per 100 000 during the intervention (13% increase, −30 to 83; p=0·63)
Datiko et al (2017)22	Ethiopia, remote rural	Community mobilisation, door to door, community health workers collecting and transporting sputum	Sputum smear if symptoms	Laboratory strengthening, LTBI treatment of child contacts, contact tracing	Microbiologically confirmed	72·4	107·3	1·48	79·1	85·0	1·08	1·3	In the intervention region during the baseline period, there were 64 (95% CI 62.5–65.8) sputum smear-positive cases and 102 (99.1–105.8) cases of all-form tuberculosis per 100 000 population per year, increasing to 127 cases of smear-positive and 177 cases of all-form tuberculosis per 100 000 population per year in the endline period. In the control region, 86 cases of smear-positive and 185 cases of all-form tuberculosis per 100 000 population per year were reported in the endline period, which was similar to baseline (p>0.1)
Aye et al (2018)23	Myanmar, informal urban (and neighbourhood contacts)	Door to door for neighbourhood contacts, community mobilisation for others; volunteers collecting sputum	Sputum tests if symptoms (mainly sputum smear, Xpert for people with HIV or retreatment); chest x-ray and clinical assessment if no sputum produced	Financial incentives for volunteers, contact tracing	All types	142	148·2	1·04	239·0	195·3	0·82	1·28	Average difference in CNR between intervention and control townships declined by 50·9 cases per 100 000 population per year (95% CI −10 to 112) during the intervention period, but this finding was not statistically significant (p>0·05)‡
Vyas et al (2019)24	India, indigenous group	Door to door, community health workers collecting and transporting sputum	Sputum smear if symptoms	Financial incentives for volunteers	Microbiologically confirmed	90·7	166·7	1·84	83·9	79·3	0·95	1·94	The tuberculosis notification trend in the intervention area in the baseline period was slightly negative; regression analysis showed increases compared with expected notification rates of 89·4% for smear positive cases and 90·8% for all types of tuberculosis in the endline period; in the control area, smear-positive notifications decreased slightly (−5·5%)
Chen et al (2019)25	China, general population	Door to door, community health workers collecting and transporting sputum	Chest x-ray if symptoms or in high-risk group. Sputum smear if symptoms or abnormal chest x-ray	None	All types	78·5	67·7	0·86	79·0	62·6	0·79	1·01	No significant difference found between the cumulative incidence proportion for ACF (67·7 per 100 000 population) and the prevalence for passive case-finding (62·6 per 100 000 population) during the intervention period; authors report CNR ratio intervention vs control for each year separately§
Shewade et al (2019)26	India, indigenous populations plus informal urban	Door to door, community mobilisation, volunteers collecting and transporting sputum	Sputum smear if symptoms	Financial incentives for volunteers, engagement with non-governmental organisations	Microbiologically confirmed	15·8	15·3	0·97	14·1	11·8	0·84	1·16	After the active case-finding intervention was introduced, sputum-positive CNR per 100 000 population increased, with a β coefficient of 1·3 (95% CI 0·6–2·0)

The control intervention was usual case-finding in all studies. CNR=case notification rate. ACF=active case-finding. TST=tuberculin skin test. LTBI=latent tuberculosis infection. IDP camp=camp for internally displaced people. NA=not applicable.

^*The study does not specify whether this p value was adjusted for the presence of clustering.

^†No population estimate was provided, so it was not possible to calculate CNRs; we calculated CNR ratios from numbers of tuberculosis diagnoses, assuming that the underlying population denominator remained the same.

^‡The value quoted (50·9) is a coefficient from a general estimating equation which indicates the average change in the difference in tuberculosis notification rates per year between intervention townships and non-intervention townships in the intervention and control period (ie, an interaction term between intervention and control townships and intervention and control time periods after adjusting for secular trends); the p value given for this coefficient is 0·11.

^§For 2013, the CNR ratio comparing intervention area to control area is 1·7 (95% CI 1·2–2·5), for 2014 it is 1·3 (0·8–1·9), and for 2015 is 0·2 (0·08–0·6); the study does not state whether these findings are adjusted for clustering or not.