A total of 2483 references were retrieved, from which 52 evaluated interventions working with private for-profit providers were identified. The majority concerned training (26) and social marketing (14), with the remainder evaluating contracting-out (3), franchising (6), regulation (2) and accreditation (1) [see Additional file 1].
Social marketing, vouchers and pre-packaging of drugs
These interventions were grouped together because of their significant overlap. Fourteen evaluated interventions were identified, eleven relating to social marketing one also including a voucher scheme, and 2 including prepackaged treatments. A further 3 interventions focused purely on provision of vouchers.
Of the social marketing interventions, six involved condoms and/or other family planning commodities, and one each involved oral rehydration therapy (ORT), iron supplements, insecticide treated nets (ITNs), STI treatment and malaria/acute respiratory infection (ARI) treatment. All showed significant increases in utilization of programme commodities and services, though of differing magnitudes across interventions. For example, social marketing increased condom use among women in urban Cameroon from 58% to 76% , coverage of iron-folic acid supplementation from 6% to 99% in non-pregnant Filipino women , and ITN coverage in rural Tanzanian children under 2 years from 10% to 61% .
The contraceptive social marketing programmes were implemented in urban settings and the majority targeted adolescent groups. A number of interventions included peer educators as well as distribution of commodities through retail outlets (Soweto, Horizon Jeunes, Tsa Banana, My Future First).
The social marketing programme for ORT in a rural Kenyan district included sales of flavoured ORT sachets through shops and a mass communication and education campaign . In the Philippines, social marketing of iron and folic acid supplements in rural municipalities combined free distribution of iron-folic acid supplementation targeted to pregnant women through the public health system and sales of iron supplements to non-pregnant women by village health workers (VHWs) and drug outlets, as well as training VHWs and school teachers in counselling and information, education and communication (IEC) .
The Kilombero and Ulanga Insecticide-Treated Net project (KINET) in Tanzania distributed branded ITNs and net treatment kits through retail outlets, and included a comprehensive IEC campaign [19, 30, 32–34]. Vouchers reducing the price of nets at retail shops by 17% were distributed to pregnant women and mothers of under-five children attending public clinics.
Vouchers for free nets and treatment from retail outlets were used alongside measles vaccination in a campaign approach in an urban Zambian district . Two other voucher schemes for sexual and reproductive health care (SRHC) were implemented in Nicaragua. The first targeted adolescent girls at various sites, entitling the recipient to free SRHC at NGO, public and for-profit clinics [36, 37]. The second scheme provided free STI care at private for-profit and NGO clinics for sex workers and their clients .
Two final interventions combined social marketing with prepackaged treatments – for STI treatment for male urethritis in Uganda  and prepackaged treatment for childhood malaria and acute respiratory infections in Nigeria .
Of the 14 interventions, data on the average SES of the recipient community was provided for only one, the KINET ITN project. This clearly benefited a generally poor population, with an estimated median household monthly expenditure of $77–96 in 1997 . For 12 interventions, only general information on the urban/rural settings was available, and for 1 intervention there was no information aside from location.
Data on the distribution of benefits across socioeconomic groups was provided for 2 of these interventions. The effect of KINET on the socioeconomic distribution of ITNs was documented in three studies. The first demonstrated positive changes in the ratio of net ownership in the lowest to the highest SES quintiles (the equity ratio) from 0.3 in 1997 to 0.6 in 2000 . The second reported a significantly greater increase (of 51%) in household net ownership among the poorest income quartile in the social marketing area compared to 32% in the control area. Similar information for households in village peripheries (likely to be poorer) were 68% and 27% . The third study assessed the socioeconomic distribution of vouchers and, in contrast to the other two, found that none of the households in the lowest quintile had used a voucher towards the purchase of a net compared to 8% in the highest quintile . This may be due to the low share of vouchers in total net sales, itself reflecting low knowledge among mothers of vouchers, with only 28% of women having ever heard of the voucher scheme .
Poorer groups also benefited in relative terms in urban Zambia, where the equity ratio for ITN coverage increased from 0.66 to 1.19, with no statistically significant association between wealth and ownership found post-intervention .
Of the two evaluated interventions identified, one concerned banning a drug and its combination products in Nepal and one assessed a regulatory intervention to improve quality of pharmacy services in Lao P.D.R.
The pharmaceutical ban prohibited the export, import, local production, transportation, storage, sale and distribution of Analgin (an analgesic and antipyretic drug) and its combination products . As a result, the proportion of retail outlets with Analgin decreased from 96.5% at baseline to 21.2% five months after the intervention and 0% sixteen months after .
The regulatory intervention in Lao P.D.R involved intensive supervision of the quality of pharmacy services, applying sanctions when rules were violated, and providing up-to-date regulatory documents and information about particular areas needing improvements . The study compared districts with intensified regulation with normal, control districts. Whilst it could not be established that the intensive intervention had a greater effect than routine regulation, moderate but significant improvements in quality were observed in all districts, with mean availability of essential materials increasing by 34% and mean order in the pharmacy (including the presence of advertisements, and whether drugs were stored in their original packaging away from sunlight) increasing by 19% .
For these two interventions, no SES information was provided about the recipient populations.
Training was by far the most evaluated intervention, with 26 interventions covering different types of private providers: 4 targeted private doctors, 2 private midwives, 8 private pharmacy workers, 6 drug retailers and 6 a mix of provider types. Training interventions aimed to improve the quality of treatment of a range of different conditions. Seven interventions focused on treatment of childhood illness (use of integrated management of childhood illness (IMCI) guidelines, treatment of ARI or diarrhea); 5 addressed quality of STI treatment; 5 the quality of family planning or reproductive health services; 4 malaria treatment; and the remaining studies addressed other communicable diseases (e.g. ARIs), or multiple diseases (e.g. "6 common illnesses").
Most interventions produced positive results for at least some outcome indicators. For instance, a study of the Ghanaian intervention to improve STI management at pharmacies, which evaluated outcomes using simulated clients, found that when offered treatment, 38% of simulated clients received appropriate oral medication at intervention pharmacies compared with 18% at control pharmacies. Counseling about partner notification was 40% in intervention pharmacies compared with 21% in control pharmacies, though no recommendation to use a condom was given at intervention pharmacies compared with 13% at control pharmacies . Generally positive overall results were also observed for training programmes with non-pharmacy retailers. For example, the proportion of Nigerian patent medicine vendors recommending a correct drug dose for malaria increased from 9% in 2003 to 53% in 2004 , and in Kenya the proportion of antimalarial sales with adequate dosage increased from 32% in 1996 to 83% 3 months after training and to 90% 6 months after training .
Of the 26 interventions, data on average SES of the recipient population was provided for only one. Training of private practitioners in Pakistan clearly benefited a generally poor population, with an estimated median monthly household income of $48–72 and $48–61 in two communities . General information on rural/urban settings was provided for 18 interventions and no information for 7. No studies provided any evidence on the distribution of benefits by relative SES.
Six interventions were identified, "Green Star" and "Green Key" in Pakistan, "Ray of Hope" in Ethiopia, "Janani" in Bihar State, India, "Sewa" in Nepal and "Top Reseau" in Madagascar. Evidence of impact on utilization or quality of health services was mixed.
Effectiveness of the Pakistan, Ethiopia and Bihar interventions was documented in a single study that used exit interviews to examine client satisfaction at franchised and non-franchised outlets . Effectiveness of the "Green Star" and "Green Key" franchises implemented in Pakistan was jointly evaluated. Clients attending franchised private services were significantly more likely to report that they would return than those attending non-franchised services in Pakistan and significantly less likely in Ethiopia, with no statistically significant difference in Bihar. In all three settings there was no statistically significant difference between the franchise status of the clinic and perceptions of quality (that the service was better than others available) or in citing affordability as a preferred feature of the service .
The Nepali study examined client satisfaction with quality of care . Clients at intervention clinics 'very satisfied' with cleanliness increased from 37% to 65%, and with the availability of essential equipment from 35% to 62% . Clients were also reported to be more satisfied with the range of services offered in the intervention clinics (40% to 71%) and with privacy (38% to 72%) .
The Top Réseau study reported that coverage of modern contraceptives was higher for women with high exposure to the intervention (have accessed a franchised clinic and have been exposed to the IEC activities) than those with low or medium exposure .
Data on average SES of the recipient population was provided for one of the six interventions. The Nepali franchise network clearly benefited a generally poor population, with an estimated income per capita of $125 .
Evidence on the socioeconomic distribution of benefits within the recipient community was provided for the franchise networks in Pakistan, Ethiopia and Bihar State. The Green Star and Green Key interventions in urban Pakistan did not benefit relatively poor groups. Clients with income of $101– $250/month and with income greater than $251 were more likely to use franchised services than those earning less than $60/month. Clients with at least secondary schooling were also more likely to use franchised services compared to illiterate clients .
Evidence for the Ray of Hope intervention in Ethiopia was mixed. Clients with income of $101–$250/month were less likely to attend franchised services than those earning less than $60/month. However clients with primary education were more likely to attend franchised services compared to clients with the least education .
The evidence from Bihar was also mixed. There was no statistically significant association between attending franchised services and monthly household income . Clients with no education were more likely to attend franchised services compared to clients with education .
Only general SES information was provided for the Top Réseau study.
One intervention was identified, a network of accredited drug dispensing outlets (ADDO) implemented in rural and peri-urban Tanzania. The accreditation process, managed by the Tanzania Food Drug Authority (TFDA), aimed to improve access to affordable and quality medicines and pharmaceutical services through training and supervision of outlet dispensing staff, outlet inspections, marketing and public education . The proportion of unregistered drugs decreased in both intervention and control areas, from 26% to 2% in the former, and from 29% to 10% in the latter .
Only general information on the rural and peri-urban status of the recipient populations was available, and no information was provided about ADDO customers' SES.
Three evaluated interventions were identified, of which one related to contracting-out hospital services in South Africa and two related to contracting-out primary health care services in South Africa and Lesotho.
Contracting-out of district hospitals to private-for-profit management was implemented in rural South Africa. The quality of care provided by three contracted hospitals was compared with that of three, paired public hospitals . Public hospitals had better structural quality of care but contracted hospitals had better quality of nursing care in maternity and medical/surgical wards than public hospitals, similar nursing management quality, and overall, higher total nursing quality. No statistically significant differences in perinatal and maternal mortality rates were found between contracted and public hospitals .
General practitioners have for long been contracted on a part-time basis to provide primary care in rural towns in South Africa . A quality of care study showed that patients with hypertension were less likely to have their blood pressure recorded when they sought care at contracted practices than at public health facilities.
Primary care services, drugs, laboratory tests and X-rays were provided in Lesotho to workers of a construction company and to local communities through a contract with a commercial medical company . Overall, structural quality was similar between contracted and public providers. However, 37% of STI cases were treated correctly by contracted providers compared with 59% and 96% of cases treated in "large" and "small" public health facilities respectively.
Of the 3 evaluated interventions, data on the average SES of the recipient community was provided for 2. The individual GP and the company contract interventions clearly benefited generally poor users given that approximately 65% to 78% of them had an estimated household monthly income of less than $66 .
For the intervention to contract out district hospitals, rural location was the only information about SES provided.