A predictive model for retention in specialized HIV/AIDS care

Miranda, Wenislayne de Araújo; Medeiros, Leidyanny Barbosa de; Nascimento, João Agnaldo do; Ribeiro, Kátia Suely Queiroz Silva; Nogueira, Jordana de Almeida; Leadebal, Oriana Deyze Correia Paiva

Acessibilidade / Reportar erro

Brasil

Cadernos de Saúde Pública

Español English

Brasil

Español English

sumário « anterior atual seguinte »

Sumário

ARTICLE • Cad. Saúde Pública 34 (10) • 2018 • https://doi.org/10.1590/0102-311X00209416 copy

A predictive model for retention in specialized HIV/AIDS care

Authorship SCIMAGO INSTITUTIONS RANKINGS

The establishment of universal targets for HIV/AIDS control and the implementation of treatment as prevention reinforce the need for on-going clinical follow-up of persons living with HIV/AIDS as an essential element of their care, where retention in care is both a need and a challenge. This study aimed to create a predictive model for retention of persons living with HIV/AIDS in health care. A decision tree statistical model was created, based on sociodemographic, clinical, and health behavior variables, identified in a database with information from 260 persons with HIV/AIDS, enrolled in a specialized treatment service. The model enabled the identification of nine variables with significant information gains in relation to the outcome variable, probable retention in health care, and the development of 24 decision rules, giving rise to a decision tree with 80.4% correct answers, which can help identify possible strategies to optimize retention and contribute to achieving the proposed targets for confronting the epidemic in the coming years.

Keywords:
Acquired Immunodeficiency Syndrome; Highly Active Antiretroviral Therapy; Delivery of Health Care; Decision Threes

Classification of sample (real)	Classification of decision tree	Total
	Retention in care	Probable non-retention	Probable retention
	Non-retention	43 *	39	82
	Retention	12	166 *	178
	Total	55	205	260
Total correct answers	80.4%

Model’s quality		95%CI
Accuracy	80.4%	75.6; 85.2
Sensitivity	78.2%	67.3; 89.1
Specificity	81.0%	75.6; 86.3
Pretest probability	21.2%	16.2; 26.1
Positive predictive value	52.4%	41.6; 63.2
Negative predictive value	93.3%	89.6; 96.9
Positive likelihood ratio	4.1	-
Negative likelihood ratio	0.3	-

Attribute	Decision
1. IF patient is not using ART	THEN retention in care is unlikely
2. IF patient is using ART	THEN it will be necessary to verify last viral load
3. IF patient is using ART and the last viral load is undetectable	THEN retention in care is likely
4. IF patient is using ART and there has been no viral load test	THEN retention in care is unlikely
5. IF patient is using ART and the last viral load is detectable	THEN it will be necessary to verify patient’s age
6. IF patient is using ART and the last viral load is detectable and the patient is less than 20 years old	THEN retention in care is unlikely
7. IF patient is using ART and the last viral load is detectable and the patient is 40-59 years old	THEN retention in care is likely
8. IF patient is using ART and the last viral load is detectable and the patient is 60 years or older	THEN retention in care is likely
9. IF patient is using ART and the last viral load is detectable and the patient is 20-39 years old	THEN it will be necessary to verify the patient’s sexual orientation
10. IF patient is using ART and the last viral load is detectable, the patient is 20-39 years old and homosexual	THEN it will be necessary to verify whether the patient uses alcoholic beverages
11. IF patient is using ART and the last viral load is detectable, the patient is 20-39 years old, homosexual, and uses alcoholic beverages	THEN retention in care is unlikely
12. IF patient is using ART and the last viral load is detectable, patient is 20-39 years old, homosexual and does not use alcoholic beverages	THEN retention in care is likely
13. IF patient is using ART and the last viral load is detectable, patient is 20-39 years old and bisexual	THEN retention in care is likely
14. IF patient is using ART and the last viral load is detectable, patient is 20-39 years old and there is no information on sexual orientation	THEN retention in care is likely
15. IF patient is using ART and the last viral load is detectable, patient is 20-39 years old and heterosexual	THEN it will be necessary to verify whether the patient has been hospitalized in the service
16. IF patient is using ART and the last viral load is detectable, patient is 20-39 years old, heterosexual and has been hospitalized in the service	THEN it will be necessary to verify the patient’s marital status
17. IF patient is using ART and the last viral load is detectable, patient is 20-39 years old, heterosexual, has been hospitalized in the service, and single marital status	THEN retention in care is likely
18. IF patient is using ART and the last viral load is detectable, patient is 20-39 years old, heterosexual, has been hospitalized in the service, and marital status is separated/widow(er)	THEN retention in care is likely
19. IF patient is using ART and the last viral load is detectable, patient is 20-39 years old, heterosexual, has been hospitalized in the service, marital status is married or in stable union	THEN it will be necessary to verify whether the patient lives in the same municipality as the service
20. IF patient is using ART and the last viral load is detectable, patient is 20-39 years old, heterosexual, has been hospitalized in the service, marital status is married or in a stable union, and lives in the same municipality as the service	THEN retention in care is likely
21. IF patient is using ART and the last viral load is detectable, the patient is 20 to 39 years old, heterosexual, has been hospitalized in the service, marital status is married or in a stable union, and does not live in the same municipality as the service	THEN retention in care is unlikely
22. IF patient is using ART and has detectable viral load, patient is 20-39 years old, heterosexual, and has never been hospitalized in the service	THEN it will be necessary to verify how many pills per day the patient is taking
23. IF patient is using ART and has detectable viral load, patient is 20-39 years old, heterosexual, has never been hospitalized in the service, and takes fewer than 4 pills/day in ART	THEN retention in care is unlikely
24. IF patient is using ART and has detectable viral load, patient is 20-39 years old, heterosexual, never been hospitalized in the service, and takes 4 or more pills/day	THEN retention in care is likely

Clinical case

Application of the model

Male patient, 19 years old, homosexual. At time of reporting, patient presented some clinical manifestations, was in antiretroviral therapy, taking 2 pills a day, and last viral load, 3 months ago, was detectable. No previous hospitalizations in the service. Given this context, what will the patient’s behavior be in follow-up/retention in the service/network of care?

The decision variables in the tree are arranged hierarchically, starting from the root node (first variable). In this model it is represented by the variable ART use. Starting interpretation from rule 2: IF patient is using ART, THEN it will be necessary to verify the last viral load. Since the last viral load was detectable, analyze rule 5: IF patient is using ART and the last viral load is detectable, THEN it will be necessary to verify the patient’s age. Since the patient is 19 years old, rule 6 will be used: IF patient is using ART and the last viral load is detectable and the patient is less than 20 years old, THEN retention in care is unlikely.

Escola Nacional de Saúde Pública Sergio Arouca, Fundação Oswaldo Cruz Rua Leopoldo Bulhões, 1480 , 21041-210 Rio de Janeiro RJ Brazil, Tel.:+55 21 2598-2511, Fax: +55 21 2598-2737 / +55 21 2598-2514 - Rio de Janeiro - RJ - Brazil
E-mail: cadernos@ensp.fiocruz.br

Acompanhe os números deste periódico no seu leitor de RSS

[1] Correspondence W. A. Miranda Universidade Federal da Paraíba. Campus I, Castelo Branco, João Pessoa, PB 58051-900, Brasil. wenislayne@hotmail.com