HIV: California’s 2018 criminalization reform and testing among those reporting risk behavior

Abstract

HIV criminalization laws may discourage HIV testing. We tested whether California’s 2018 HIV criminalization law reform increased the likelihood of past-year HIV testing compared to Nevada, which did not reform its HIV criminalization law. We fitted two difference-in-differences logistic regression models: one for all respondents reporting behaviors that increase the chances of getting or transmitting HIV, and one for male respondents reporting these behaviors. All analyses accounted for the complex survey design of BRFSS. HIV criminalization reform was significantly associated with an increased likelihood of past-year HIV testing. After reform, the predicted marginal probability of past-year HIV testing increased by six percentage points. By comparison, probabilities of a past-year HIV test decreased in Nevada. HIV criminalization law reform may increase the likelihood of getting tested by individuals who engage in behaviors that increase the chances of getting or transmitting HIV.

Introduction

HIV testing is one of the most important strategies for reducing HIV acquisition. While individuals with undiagnosed HIV make up 13.3% of persons with HIV in the U.S. [1], this group accounts for nearly 40% of all cases of HIV transmission [2]. Thus, widespread HIV screening has the potential to reduce HIV transmission by increasing early HIV diagnosis, reducing the number of persons with HIV who are unaware of their HIV status and facilitating linkage to care. For this reason, the U.S. Centers for Disease Control and Prevention (CDC) recommends that everyone between the ages of 13 and 64 get tested for HIV at least once, and that those with risk factors for HIV be tested more frequently [3].

HIV criminalization laws, defined as HIV-specific laws that criminalize or control behaviors that can potentially expose another person to HIV [4], have long been viewed as a structural barrier to HIV testing [5]. It has been widely hypothesized that HIV criminalization laws may discourage HIV testing [6]. However, there have been few quantitative analyses that measure the effects of HIV criminalization laws, including only one empirical study to date on associations between state HIV criminalization laws and HIV testing [7].

California’s recent (2018) reform of its HIV criminalization law provides an opportunity for a “natural experiment” to estimate the independent effect of legal reform on individual likelihood of recent HIV testing using a difference-in-differences analysis. Under California’s previous HIV criminalization statute, for an individual with HIV, having unprotected anal or vaginal sex with a partner without first disclosing his or her HIV status was a felony, even if transmission did not occur. Donation of blood, organs, tissue, or semen was also a felony. California significantly modified its HIV exposure statute in 2017 (going into effect on January 1, 2018), removing the explicit reference to HIV, making it a misdemeanor to transmit (rather than merely expose) an infectious disease, requiring specific intent to be demonstrated, and reducing the maximum possible sentence to six months [8]. Nevada, which had an HIV criminalization statute that remained unchanged during the analytic time period, was chosen as the control state because of its geographic proximity and similar demographic makeup.

To test the hypothesis that California’s HIV criminalization reform increased the likelihood of individuals at elevated risk for HIV reporting an HIV test in the previous 12 months, we conducted a difference-in-differences analysis comparing California and Nevada survey respondents who reported behaviors that increased the chances of getting or transmitting HIV and testing history from 2016 to 2020.

Data and methods

Data sources

HIV testing history, self-reported behaviors that increase the chances of getting or transmitting HIV, county of residence, and individual-level demographic and socioeconomic covariates (age, sex, race, ethnicity, marital status, poverty, education, and employment) were obtained from the CDC’s Behavioral Risk Factor Surveillance System (BRFSS), a system of telephone surveys that collects data in all 50 states and the District of Columbia. A respondent was classified as someone who self-reported behaviors that increase the chances of getting or transmitting HIV if they answered “Yes” to the following question: “I am going to read you a list. When I am done, please tell me if any of the situations apply to you. You do not need to tell me which one. You have injected any drug other than those prescribed for you in the past year. You have been treated for a sexually transmitted disease or STD in the past year. You have given or received money or drugs in exchange for sex in the past year. You had anal sex without a condom in the past year. You had four or more sex partners in the past year. Do any of these situations apply to you?” Data on HIV criminalization laws by state were obtained from the U.S. Centers for Disease Control and Prevention ⁴ and the Center for HIV Law and Policy [9]. The urbanization level for each county (on an ordinal scale of 1–6) was taken from the National Center for Health Statistics Urban–Rural Classification Scheme for Counties [10]. County-level HIV prevalence rates per 100 population were obtained from AIDSVu [11].

Study sample and descriptive analysis

This analysis included participants who reported behavior that increase the chances of getting or transmitting HIV (“Do Any [HIV-related] High Risk Situations Apply”) with non-missing data (a response of “Yes” or “No” and, for those who responded “Yes,” a valid date) for HIV testing history in the previous year (“Including fluid testing from your mouth, but not including tests you may have had for blood donation, have you ever been tested for HIV?”) from the 2016–2020 BRFSS for California and Nevada. The amount of time since the respondent’s most recent HIV test was calculated as the number of months between the interview date and the most recent HIV test. If the number of months was 12 or fewer, the respondent was coded as having reported an HIV test in the past year.

Descriptive statistics were generated to characterize the population represented by the survey sample. Variables included individual respondent sex at birth, age group, race, Hispanic/Latino ethnicity, marital status, education level, poverty status, and employment status, as well as urban–rural classification for county of residence.

Logistic regression models and predicted marginal probabilities

Two multivariate logistic regression models were fitted: one with all BRFSS participants with HIV testing history information who reported a behavior that increases the chances of getting or transmitting HIV in the previous year (n = 3,294), and one with only males. A significance level of α = 0.1 was chosen a priori. We designated the post-reform period as beginning with the 2019 survey year, so that all respondents in the post-period who were coded as having reported an HIV test in the previous year would have their test fall after the legal reform went into effect on January 1, 2018. The years prior (2016–2018) were all coded as pre-reform. Terms for the state of California (during the pre-reform period) and for the post-reform period (years 2019 and 2020) in Nevada, which served as the reference group, as well as an interaction between the two (representing California during the post-reform period, to estimate the independent effect of legal reform), were included in both models. Weighted percentages of those who reported testing for HIV in the previous 12 months were plotted for both states for each year in Fig. 1, to check the parallel trends assumption.

Fig. 1.

Weighted percentages of respondents reporting a past-year HIV test in California and Nevada. Secular trends were comparable in the pre-reform period (2016–2018), satisfying the parallel trends assumption, but diverged in the post-reform period (2019–2020)

Predicted marginal probabilities were generated for California and Nevada for the pre- and post-reform periods. Survey year was included to account for secular trends. All analyses were adjusted for the BRFSS complex survey design and accounted for the pooling of multiple survey cycles and sampling design effects using the SURVEY family of SAS procedures. The analyses were conducted using SAS 9.4 [14] and SUDAAN 11.0.3 [15].

Results

Study sample

Among all respondents from the two states (n = 65,670), 3,652 reported a behavior that increased the chances of getting or transmitting HIV in the previous year. Of those, 3,353 respondents provided information on HIV testing history that included at least the year of the most recent HIV test, which allowed us to determine whether the respondent had been tested for HIV in previous year. The final analytic sample included 3,294 participants.

Descriptive analysis

Table 1 shows descriptive statistics for the study sample, comparing weighted percentages between California and Nevada.

Table 1.

Demographic and socioeconomic characteristics of the study sample: Behavioral Risk Factor Surveillance System, 2016–2019

	Full analytic sample (n = 3,294)					Male sex at birth subsample (n = 2,086)
	California		Nevada			California		Nevada
	n a	%	n a	%	p b	n a	%	n a	%	p b
Sex at birth
Female	1,655	61.3	431	59	0.42
Male	934	38.7	274	41
Race
White	1,661	59.9	465	53.8	< 0.01	1,032	58.5	290	58.1	0.52
Black or African American	199	8.5	63	14.5		124	9.1	35	11.6
American Indian or Alaska Native	86	3.0	17	3		63	3.7	11	2.7
Asian	187	11.7	23	8.9		127	10.4	14	8.6
Multiracial or other	372	12.6	122	17.2		255	13.9	72	16.4
Unknown race	84	4.3	15	2.6		54	4.4	9	2.5
Hispanic/Latino ethnicity
No	1,602	57.9	528	70.7	< 0.01	1,010	55.7	314	67.9	< 0.01
Yes	987	42.1	177	29.3		645	44.3	117	32.1
Age group
18–24 years	712	32.6	170	28.5	0.51	429	29.0	97	25.8	0.77
25–34 years	847	33.3	244	35.1		551	35.5	151	36.6
35–44 years	445	16.0	118	18.3		281	16.3	63	17.6
45–54 years	307	10.0	83	11.3		192	9.7	57	11.7
55–64 years	176	5.1	52	4.3		130	6.7	38	5.3
65 + years	102	3.0	38	2.5		72	2.7	25	3.0
Married
No	1,794	68.7	490	69	0.89	1,172	70.9	310	71.9	0.79
Yes	795	31.3	215	31		483	29.1	121	28.1
Education
Did not graduate high school	250	13.1	67	12	0.38	166	13.6	36	10.4	0.43
Graduated high school	579	26.5	202	30.2		401	29.1	130	29.7
Some college	1,760	60.4	437	57.8		1,088	57.3	265	59.9
Employed
No	770	32.7	208	26.6	0.02	418	27.1	111	22.0	0.10
Yes	1,819	67.3	497	73.4		1,237	72.9	320	78.0
Below federal poverty level
No	1,910	70.9	535	73.4	< 0.01	1,276	74.3	348	81.0	< 0.01
Yes	479	20.2	80	11.1		264	17.7	34	5.7
Unknown	200	8.9	90	15.4		115	8.0	49	13.3
County urban–rural classification
Large central metro	1,652	64.9	312	74.5	- c	1,102	66.3	191	75.2	- c
Large fringe metro	326	12.4	-	-		186	11.4	-	-
Medium metro	453	17.5	250	17		275	17.5	155	16.4
Small metro	82	3.0	27	1.3		46	2.5	19	1.6
Micropolitan	52	1.4	104	6.5		32	1.3	61	6.3
Non-core	24	0.8	12	0.8		14	0.9	5	0.4

^aWeighted percentages may differ slightly from unweighted n’s

^b The p value is for the Rao-Scott chi-square test

^cThe statistical test could not be conducted for county urbanicity because there were no observations in the “large fringe metro” category for Nevada

Notable differences for the full analytic sample include the racial and ethnic distributions between states. Additionally, the weighted percentages of those above and below the federal poverty level, as well as those with unknown poverty status, were significantly different in California vs. Nevada. Both states show comparable age, marital status, education level, and county urbanicity distributions.

While the weighted percentages of each racial group were more similar among males than among the overall population, California had a larger weighted percentage of Hispanic/Latino individuals than Nevada (44.3% vs. 32.1%, p < 0.01). Both states had similar age, marital status, educational attainment, and county urbanicity distributions among those who reported male sex at birth represented by the sample. California had a lower weighted percentage of those employed (p = 0.1) and living above the federal poverty line (p < 0.01) than Nevada.

Model results

The parameter estimates for the models fitted with the full analytic sample and for those reporting male sex at birth are shown in Table 2.

Table 2.

Parameter estimates for the effect of California’s HIV criminalization reform on past-year HIV testing likelihood among individuals who report HIV risk behavior

Parameter	Full analytic sample					Male sex at birth subsample
	Odds Ratio	90% CI		p		Odds Ratio	90% CI		p
Location
California	1.12	1.00	1.26	0.09	*	1.16	1.00	1.34	0.09	*
Nevada	(Ref.)					(Ref.)
Time period
Post-2018	1.06	0.90	1.26	0.56		0.97	0.78	1.2	0.80
Pre-2018	(Ref.)					(Ref.)
HIV criminalization reform a	1.13	1.01	1.26	0.07	*	1.18	1.03	1.36	0.05	*
Age	0.99	0.98	0.99	<0.01	*	1.00	0.99	1.01	0.52
Sex at birth
Male	0.95	0.79	1.14	0.62
Female	(Ref.)
Race
Black	1.55	1.16	2.07	<0.01	*	1.52	1.05	2.21	0.02	*
American Indian/Alaska native	0.98	0.6	1.57	0.84		0.91	0.5	1.65	0.82
Asian	0.59	0.4	0.87	0.03	*	0.72	0.44	1.18	0.25
Multiracial or other race	0.93	0.71	1.22	0.99		0.96	0.69	1.34	0.95
Unknown race	0.77	0.48	1.25	0.45		0.87	0.49	1.55	0.72
White	(Ref.)					(Ref.)
Hispanic/Latino ethnicity
Yes	1.2	0.98	1.47	0.15		1.22	0.95	1.59	0.20
No	(Ref.)					(Ref.)
Marital status
Married	0.6	0.49	0.73	<0.01	*	0.64	0.5	0.81	<0.01	*
Unmarried	(Ref.)					(Ref.)
Educational attainment
Did not finish HS	0.39	0.26	0.56	<0.01	*	0.30	0.19	0.48	<0.01	*
HS diploma	0.64	0.51	0.8	0.87		0.67	0.51	0.89	0.30
Some college	(Ref.)					(Ref.)
Employment
Employed	0.97	0.79	1.2	0.83		0.97	0.73	1.28	0.85
Unemployed	(Ref.)					(Ref.)
Below federal poverty level
Yes	1.19	0.93	1.53	0.19		0.97	0.69	1.37	0.82
No	(Ref.)					(Ref.)
Unknown	0.92	0.64	1.31	0.42		1.06	0.66	1.71	0.82
County urban–rural classification
Large central metro	1.07	0.46	2.49	0.23		0.78	0.26	2.38	0.07
Large fringe metro	1.18	0.5	2.83	0.11		0.54	0.17	1.68	0.84
Medium metro	1.03	0.44	2.44	0.39		0.86	0.28	2.65	0.04	*
Small metro	0.82	0.31	2.16	0.74		0.47	0.13	1.66	0.61
Micropolitan	0.49	0.18	1.31	0.04	*	0.19	0.05	0.66	<0.01	*
Non-core	(Ref.)					(Ref.)
County HIV prevalence rate	2.28	1.61	3.24	<0.01	*	2.21	1.47	3.34	<0.01	*
Year	0.97	0.87	1.09	0.70		1.01	0.87	1.17	0.92

^*Significant at α = 0.1

^aThe interaction term for the state of California and the post-2018 period, representing the reform of California’s HIV criminalization law, is a ratio of odds ratios

Compared with respondents living in Nevada, living in California was associated with an increased likelihood of reporting an HIV test in past year among those reporting a behavior that increases the chances of getting or transmitting HIV. The time period after HIV criminalization reform was not significant. California’s HIV criminalization reform, represented in the model by the interaction term between the state of California and the post-reform (post-2018) period, was independently associated with an increased likelihood of getting tested for HIV in the previous year among those reporting a behavior that increases the chances of getting or transmitting HIV, and the effect was statistically significant (p = 0.07). Other factors associated with an increased likelihood of reporting an HIV test in the past year included Black or African American race and higher county-level HIV prevalence. Asian race, higher age, being married, not having a high school diploma, and living in a micropolitan county were all associated with decreased likelihood of reporting a past-year HIV test among the full analytic sample. Hispanic/Latino ethnicity, employment and poverty status, and living in a large, medium, or small metro county were not associated with recent HIV testing likelihood.

For males, living in California was similarly associated with an increased likelihood of reporting an HIV test in past year compared to those living in Nevada. The period after HIV criminalization reform was not significant. California’s HIV criminalization reform, represented in the model by the interaction term between the state of California and the post-reform (post-2018) period, was also associated with an increased likelihood of getting tested for HIV in the previous year among those of male sex at birth (p = 0.05).

Table 3 shows the predicted marginal probabilities for past-year HIV testing among those reporting a behavior that increases the chances of getting or transmitting HIV in California and Nevada in the pre- and post-reform periods for the full analytic sample and for respondents who were male sex at birth.

Table 3.

Predicted marginal probabilities pre- and post-reform for past-year HIV testing history among individuals who report HIV risk behavior

	Full analytic sample	Male sex at birth subsample
California, pre-2018	0.35	0.35
California, post-2018	0.42	0.41
Nevada, pre-2018	0.36	0.36
Nevada, post-2018	0.32	0.28

For the full analytic sample, the pre-reform predicted marginal probabilities were similar for California (0.35) and Nevada (0.36). However, the post-reform probability for Nevada decreased by 0.0393 to 0.3163, while the post-reform probability for California increased by 0.0648 to 0.4176. A similar pattern was observed for the predicted marginal probabilities for male sex at birth respondents: while the pre-reform probabilities for past-year HIV testing were comparable for both states (0.3514 and 0.3600 for California and Nevada, respectively), the post-reform probability for California increased by 0.0594 (to 0.4108), while it decreased by 0.0818 for Nevada (0.2782).

Discussion

The goal of this study was to test whether California’s reform of its HIV criminalization law was associated with an increase in the individual likelihood of reporting an HIV test in the past 12 months among those who report behavior that increases the chances of getting or transmitting HIV, compared to individuals in this group in Nevada, which did not reform its law until 2021. The analysis was able to take advantage of the change in a state law by comparing it to a neighboring state, which did not reform its own HIV criminalization law, using a difference-in-differences analysis with a quasi-experimental design. The results show that, after California’s HIV criminalization law reform, there was an approximately six percentage point increase in the predicted marginal probability of a past-year HIV test in this population for both the entire analytic sample and for those reporting male sex at birth. By comparison, the predicted marginal probabilities of a past-year HIV test decreased in Nevada in both groups.

Public health professionals, lawyers, and advocates have long argued that laws criminalizing HIV are ineffective and may be counterproductive [5, 16] because they further criminalize groups that already face health inequities that increase their chances of getting HIV, including Black and Hispanic/Latino individuals, and counteract effective public health messages on HIV prevention and safer sex. Over the course of the HIV epidemic, researchers have warned that HIV criminalization laws may indirectly increase HIV transmission [7, 16] by motivating individuals targeted by such laws to avoid testing and, as a result, culpability under the law. An overwhelming number of commentators on HIV criminalization laws have noted that it is impossible to enforce HIV criminalization laws uniformly across the entire population [17, 18], and that there is evidence that HIV criminalization laws are disproportionately enforced against Black and Hispanic/Latino individuals [19, 20]. In a descriptive analysis of criminal prosecutions for HIV exposure, Lazzarini, Bray, and Burris [21] found no evidence of systemic enforcement of these laws but noted that “[w]hat seems to determine who gets prosecuted is the accident of being caught.” Research has shown that heavier policing of Black men drives them to “go into hiding,” reducing their participation in civic social life to avoid contact with “the system”–that is, law enforcement and the criminal justice system [22]. This is true of Hispanic/Latino individuals as well, who have also been shown to be significantly more likely to report having negative interactions with police, to believe that police engage in misconduct, and to live in heavily-policed neighborhoods [23]. This may contribute to Black and Hispanic/Latino men avoiding HIV testing as a means of “system avoidance” [24]. While Black and Hispanic/Latino individuals are more likely than other racial and ethnic groups to get tested for HIV [25–27], they are also more likely to be tested and diagnosed late in the course of infection. Previous studies using NHBS data [28] and nationally representative survey data [25] have found that around a third of Black and Hispanic/Latino participants have never been tested for HIV, despite its increased prevalence among these groups. This “system avoidance” that creates lower rates of HIV testing may be exacerbated in states with HIV-specific criminalization laws [7, 29]. This analysis represents an important contribution to the literature, as it provides quantitative evidence that individuals with behaviors that increase their chances of getting or transmitting HIV are less likely to report a recent HIV test in states with HIV criminalization laws, and that HIV criminalization reform can increase the likelihood of HIV testing among individuals in this group.

Limitations

This analysis is subject to several limitations. Non-response (selection) bias in the BRFSS data may influence model results. Response rates for BRFSS are relatively low, ranging from a high of 51.8% for Nevada in 2019 to a low of 31.1% for California in 2016. Additionally, response rates are consistently lower in California than in Nevada, with differences ranging from 9.2 percentage points in 2020 to 13.2 percentage points in 2017. This is particularly important in light of the populations most likely to be impacted by HIV criminalization laws, whose members have been shown to engage in “system avoidance” (as described above), as BRFSS is a government-sponsored telephone survey.

There is also the potential for information bias. Among all respondents (n = 65,670), 11.8% (n = 7,728) did not provide any information on behavior that increases the chances of getting or transmitting HIV. Among those who reported such behavior (n = 3,652), 1.2% (n = 45) are missing HIV testing information, and a further 7.0% (n = 254) who reported having ever been tested for HIV did not provide the month and year of last test. While missingness for demographic and most socioeconomic variables was low (< 4% for race, < 2% for Hispanic/Latino ethnicity, and < 1% for all other variables), poverty status could not be determined for 9.2% of respondents included in the study sample. Among male respondents (n = 32,005), 12% (n = 3,834) did not provide any information on these behaviors. Among those who reported behavior that increases the chances of getting or transmitting HIV (n = 2,309), 0.9% (n = 21) are missing HIV testing information, and a further 6.8% (n = 158) who reported having ever been tested for HIV did not provide the month and year of last test. While missingness for demographic and most socioeconomic variables was similarly low for males, poverty status could not be determined for 8.4% of respondents included in the study sample.

There is further potential for information bias due to social desirability pressures. Information about HIV testing and related risk factors is still commonly viewed with stigma, potentially influencing respondents to refuse to provide information on their HIV testing history, on their sexual or drug use behaviors, or whether they have been treated for an STI.

While California and Nevada both had HIV-specific criminalization statutes before 2018, there may have been differences in the way these laws were enforced between states that were not accounted for in the analysis. Finally, if there were any social, structural, or programmatic changes that impacted HIV testing likelihood during the post-reform period (after 2018) in one state and not the other, they could have contributed to the results we observed.

Strengths

Despite the limitations in the available data, this study has several notable strengths, including its large sample size, use of recent data, and quasi-experimental design. To date, only one other empirical analysis on the relationship between HIV criminalization laws and HIV testing has been published [7], and it uses much older data. Thus, this analysis will contribute to addressing a gap in knowledge by using more recent data to contribute to the literature examining the effects of HIV criminalization laws on HIV testing. Additionally, this is the first published analysis utilizing a design that allows for causal inference. Researchers have long used difference-in-differences analyses to draw causal inferences from observational data, particularly for policy interventions [30]. For example, this design was used to estimate the effects of Medicaid expansion on HIV testing [31].

Conclusions

HIV testing is one of the most important public health tools to prevent HIV transmission, connect those from populations with a higher incidence of HIV to other prevention interventions, and connect people with HIV to medical care. However, state HIV criminalization laws are a structural barrier to HIV testing, as they may increase stigma, exacerbate disparities, and potentially discourage individuals from testing to avoid knowing their HIV status. This study contributes to the legal and policy epidemiology literature by using a quasi-experimental design to estimate the effect of California’s reform of its HIV criminalization law on individual likelihood of reporting a recent HIV test among individuals reporting behaviors that increase the chances of getting or transmitting HIV. Our results indicate that the reform of California’s HIV criminalization law was associated with an increased likelihood of getting tested by these individuals.

Key Messages.

• Ending the HIV Epidemic in the U.S. requires addressing structural barriers to HIV prevention and care. Current scientific and medical evidence should inform state laws and practices that criminalize actions taken by people with HIV. States should consider updating or repealing outdated laws and practices.

• State HIV criminalization laws are a structural barrier to HIV testing, as they may increase stigma, exacerbate disparities, and potentially discourage individuals from testing to avoid knowing their HIV status.

• This study used a quasi-experimental design to show that California’s HIV criminalization law reform resulted in increased probability of getting tested among individuals reporting behaviors that increase the chances of getting or transmitting HIV.

Biographies

Jessica M. Keralis PhD, MPH, is an epidemiologist at U.S. Centers for Disease Control and Prevention.

Avery Bourbeau PSM, is a scientific data analyst at DLH Corporation.

Kevin P. Delaney PhD, MPH, is an epidemiologist at U.S. Centers for Disease Control and Prevention.

Shifawu Odunsi MPH, MCHES, is a public health advisor at U.S. Centers for Disease Control and Prevention.

Sheila Salvant Valentine JD, MD, MSHA, MJ, is a public health analyst at U.S. Centers for Disease Control and Prevention.

Data availability

All data used in the analysis are publicly available and cited in the Methods section.