The use of web-based diaries in sexual risk behavior research: a systematic review

Abstract

Background

An increasing number of studies have utilized the diary method, which provides quantitative event-level data about sexual encounters. Diaries are an attractive tool for sexual behavior research, yet little is known about the range of uses, methodological issues, and best practices associated with this technology.

Objectives

To conduct a systematic review of the literature regarding the use of web-based diaries in sexual risk behavior studies.

Design

Systematic review.

Data Sources

Five bibliographic databases, supplemented by references from previous reviews.

Methods

Eligible studies were published in English before August 2013, used the Internet to transmit data from collection device to study staff, and measured behaviors affecting HIV or STI transmission risk. The primary author conducted an initial screen to eliminate irrelevant articles. Both authors conducted full-text reviews to determine final articles. We abstracted data on diary methodology, validity, and reactivity (behavior change caused by diary completion).

Results

Twenty-three articles representing 15 studies were identified. Most diaries were collected daily for one month via websites, and completion was generally high (>80%). Compensation varied by study and was not associated with completion. Studies comparing diary to retrospective survey data demonstrated evidence of overreporting on retrospective tools, except for the least frequent behaviors. Most studies that assessed reactivity as a result of diary completion demonstrated some change in behavior associated with frequent monitoring.

Conclusions

Web-based diaries are an effective means of studying sexual risk behavior. More uniform reporting and further research on the extent of reactivity are needed.

INTRODUCTION

One of the greatest methodological challenges in HIV and sexually transmitted infections (STI) research is obtaining valid measures of sexual behavior. Increasingly, sexual behavior studies utilize the diary method, which provides quantitative event-level data about sexual encounters. Although potentially burdensome to participants due to the time and effort required to complete diaries, their frequent nature confers advantages over methods such as retrospective (e.g., 30-day) surveys, including improved recall.[1] Diaries typically collect daily data such as number of partners,[2] mood,[3] or STI symptoms,[4] which may be more precise than aggregate measures collected in retrospective surveys.[1,5,6] Event-level data may also allow for more accurate identification of predictors of risky behavior and HIV/STI transmission.[6]

These advantages make diaries an attractive tool for sexual behavior research. Historically, paper diaries were most common, despite several shortcomings. Compliance with the diary schedule is difficult to confirm, and participants may hoard diaries and complete multiple entries at once, increasing the potential for measurement error.[7–9] Participants in paper diary studies often must carry physical diaries with them, and may forget to complete them without frequent reminders from study staff.[1,10] Additionally, paper diaries require staff to physically enter responses into a database, leading to increased staff burden and a high potential for data entry errors given the volume of responses collected in most diary studies.[1,11]

In response to technological advances and the limitations of paper diaries, researchers have increasingly employed electronic diaries. Electronic diaries can be collected by email, website, personal digital assistant (PDA), phone, or other means of electronic data recording. This allows diaries to be time-stamped, and may be more convenient for participants.[1,10,12] Electronic diaries can implement skip patterns to reduce participant burden, identify incomplete or out-of-range responses in real-time, and reduce data entry errors.[1,10,12] The electronic format may also increase privacy.[1,12] This is an important consideration for HIV/STI diary research given that diaries may collect partners' initials or names, thereby increasing risk of confidentiality breach.

Electronic diaries, especially formats that permit remote data collection, possess enormous potential for sexual behavior research. However, little is known about the range of uses, methodological issues such as reactivity (behavior changes that result from study participation), and best practices associated with this technology. To inform the use of diaries in future studies, we conducted a systematic review of the published literature on web-based diaries used for sexual risk behavior research. Specifically, we examined the following characteristics of these studies: objectives; study populations; diary collection methods and frequency; variables measured; compensation; approaches to missing data; diary completion rates; participant acceptability and privacy; diary and survey data validity; and evidence of reactivity or behavior change as a result of diary completion.

METHODS

Search strategy and selection criteria

A systematic review of the literature was conducted in August 2013 to identify sexual risk behavior studies that used web-based diaries. We searched five databases (PubMed; EMBASE; CINAHL; MEDLINE; Cochrane Library) to identify relevant articles using the following search terms: (“diary” OR “diaries”) AND (“sexual health” OR “sexually transmitted” OR “STI” OR “STD” OR “sexual behavior” OR “HIV” OR “AIDS”). Studies published in August 2013 or earlier were eligible for inclusion.

Studies meeting the following criteria were included in our review: 1) utilized web-based diaries; 2) measured sexual risk behaviors; 3) published as a peer-reviewed article by August 2013; 4) English language. We excluded poster or presentation abstracts, case reports, letters to the editor, opinion articles, and reviews. Studies could be of any design, but the diary methodology inherently limited results to prospective studies.

We defined web-based diaries as those that used the Internet to transmit data from participants to study staff. This included website- and email-based diaries, along with cell phone or PDA diaries that transmitted data over the Internet. Studies that transferred data differently, such as through the physical exchange of memory cards, were excluded.

To identify articles most relevant to HIV/STI research, we focused on studies that measured HIV/STI incidence and/or examined risk factors for transmission. We included studies that examined condom use, whether or not they explicitly studied incidence, because of the importance of condoms in preventing transmission. Studies of sexual pleasure, contraception other than condoms, and sexual dysfunction were excluded.

Study selection and data extraction

From our database search, we generated a list of relevant citations and identified additional citations from references. The primary author scanned articles to exclude citations obviously irrelevant to our review such as studies of erectile dysfunction, sexual pleasure, or contraception other than condoms. Both authors reviewed the remaining articles, abstracted data, and integrated results to determine the final set of articles. Inconsistencies were resolved by discussion between authors. To the extent possible, we collected the following information from each study: objectives; diary medium, frequency, and collection period; study population; sample size; variables measured; participant acceptability and privacy; compensation; and how missing data were handled. Information was also abstracted on diary completion rates, validity of diary data, and reactivity. We included validity data because diaries are considered the gold standard for sexual behavior research, due to the inability of researchers to directly observe sexual behavior and the lack of confirmatory biological measures.[13–15] As a result, diaries are compared to less resource-intensive methods, such as retrospective surveys, to assess the validity of alternative methods. We also examined reactivity, the phenomenon whereby recording a behavior causes subjects to change that behavior, because its occurrence may bias observational diary studies.[1,16] We did not conduct a standardized bias or quality assessment as we were interested in studies' methods rather than conclusions. Instead, the data we collected - including completion rates, validity, and reactivity - were indicators of study quality and potential biases that assisted us in identifying optimal methodological practices.

RESULTS

Figure 1 presents a flowchart of the screening process. The initial database search yielded 1,440 unique records. An additional 245 citations did not include the specific search terms but were identified from references in reviewed articles. The primary author deemed 358 citations potentially relevant based on their titles and reviewed their abstracts to exclude ineligible studies. This yielded 35 articles for full-text review by both authors. Upon closer examination, we excluded two articles that used PDAs but did not transmit data using the Internet,[17,18] and 10 articles that did not measure sexual risk behaviors per our eligibility criteria.[19–28] This left 23 articles representing 15 studies that utilized web-based diaries to examine sexual risk behavior.[2,3,29–49]

Objectives of studies using web-based diaries

Article objectives ranged from identification of HIV/STI risk factors to methodological research. Unprotected sex was a common focus; four articles examined a range of predictors and correlates of condom use,[35,36,39,48] while others focused on specific risk factors such as substance use (n=4)[30,38,40,43] or mood (n=2).[34,41] Four assessed rates and predictors of human papillomavirus (HPV) infection in college students.[29,31,45,49] Others examined racial differences in risk behaviors of men who have sex with men (MSM),[44] online partnering among MSM,[42] and condom breakage/slippage.[46]

Several studies focused on methodological issues, including the feasibility of using diaries with specific populations[33] and high-frequency diary schedules.[3] One compared web-based diaries to text-message and paper diaries,[2] while another compared diary schedules.[33] Four assessed the validity of retrospective recall surveys.[2,32,33,37]

Methodological characteristics

Table 1 presents descriptive characteristics of the reviewed studies (see also Table 4, online supplement). The most common diary collection method was a website (n=13),[2,29–31,33–49] and the most common submission frequency was daily (n=8).[32,34–42,46,47] One study, which used cell phones to collect data, administered diaries thrice daily.[3] Seven studies collected diaries for one month.[32,34–42,46]

Table 1.

Design and Methods of Sexual Risk Behavior Studies Using Web-Based Diaries

Study	Diary Medium	Diary Submission Frequency	Diary Collection Period	Population	Sample Size*	Measures^	Compensation	Missing Data
Baer et al. (2002); Winer et al. (2006) [29,49]	Website	Biweekly	Up to 3 years	University women	82	1, 2, 3, 4, 8, 15	$2 per diary, up to $4 per month	Accounted for in model
Wilson et al. (2008); Boone et al. (2013) [30,48]	Website	Weekly	6 weeks	MSM	158	1, 2, 3, 5, 7, 9, 10	Increased from $5 to $15 per diary over course of study	Not reported
Partridge et al. (2007); Edelstein et al. (2012) [31,45]	Website	Biweekly	Up to 1.5 years	University men	240	2, 3, 4	Not reported	Missingness discussed, no methods reported
Garry et al. (2002) [32]	Email	Daily	1 month	University students	37	1, 3, 4	$50 for study participation	Not reported
Glick et al. (2013) [33]	Website	Randomized to twice weekly, weekly, biweekly, or no diary	12 months	MSM	95	1, 2, 3, 4	$1.50 per week of diaries	Imputation
Grov et al. (2010) [34]	Website	Daily	1 month	MSM	47	1, 2, 3, 4, 5, 6, 7	$1 per diary, with $10 bonus for completing ≥28 entries	Accounted for in model
Hensel et al. (2012a) [3]	Cell phone	Three times per day	3 months	STI clinic population	243	3, 4, 5, 6, 8, 12, 16	$1 per diary, plus free domestic calls, text messages, and phone-based Internet	Excluded participants with missing data
Reece et al. (2008); Hensel et al (2011) [36,46]	Website	Daily	1 month	Men	1,875	1, 2, 3, 5, 9, 11, 12	12 condoms; incentives valued up to $50–70	Missingness discussed, no methods reported
Hensel et al. (2012b) [35]	Website	Daily	1 month	MSM	3,877	1, 2, 3, 6, 9, 11, 13	Not reported	Accounted for in model
Horvath et al. (2007) [37]	Website	Daily	1 month	MSM	45	1, 2, 3, 4, 8	$50 for study participation	Not reported
Kiene et al. (2008); Kiene et al (2009) [38,39]	Website	Daily	1 month	University students	116	1, 2, 5, 6	Unspecified financial incentive prorated by diary completion; research credit	Not reported
Lim et al. (2010) [2]	Randomized to website, text message, or paper diary	Weekly	3 months	Young adults	72	1, 2, 4	$2 per paper/web or $3 per text message diary; $25 music store voucher for completing all diaries and follow-up survey	Missingness discussed, no methods reported
Mustanski (2008); Mustanski (2007a); Mustanski (2007b) [40–42]	Website	Daily	1 month	MSM	155	1, 2, 3, 4, 5, 6, 7, 10	$1 per diary, with $10 bonus for completing ≥28 entries	Excluded participants with missing data
Newcomb (2013); Newcomb and Mustanski (2013) [43,44]	Website	Weekly	3 months	MSM	143	1, 2, 3, 5, 7, 13	$60 prorated by diary completion	Excluded participants with missing data
Strachman and Impett (2009) [47]	Website	Daily	2 weeks	University students	75	1, 14	Entry into lottery for $100 for completing ≥11 diaries; research credit	Not reported

^*For studies reported in multiple articles, the largest reported sample size is presented

^{^}Measures: (1) condom use; (2) partner type; (3) type of sex; (4) partner-specific data; (5) substance use; (6) mood; (7) partner HIV status; (8) partner name or initials; (9) sexual event location; (10) method of meeting partner; (11) lubrication use; (12) contraception; (13) partner age; (14) relationship satisfaction; (15) partner circumcision; (16) non-sexual behaviors

The most frequently studied populations were MSM (n=7),[30,33–35,37,40–44,48] and university students (n=5).[29,31,32,38,39,45,47,49] Sample sizes ranged from 3,877 [35] to 37;[32] seven studies had fewer than 100 subjects.[2,29,32–34,37,47,49]

We attempted to enumerate the frequency with which diaries measured specific variables based on information provided in each manuscript. The most common measures were condom use (n=13)[2,29,30,32–44,46–49] and partner type (n=12).[2,29–31,33–46,48,49] Studies assessed partner type with varying specificity, including new vs. repeat,[2,29,31,33,37,40–45,49] casual vs. main,[2,30,34,35,38,39,48] or even further detail (i.e. girlfriend, spouse, etc.).[36,46] Diaries were also used to measure the type of sexual activity that occurred (oral, vaginal, or anal sex),[3,29–37,40–46,48,49], partner-specific rather than aggregate behaviors,[2,3,29,31–34,37,40–42,45,49] and substance use.[3,30,34,36,38–44,46,48]

Compensation for participation varied. Three studies paid participants per diary entry,[3,29,33,49] and three paid per entry with a bonus for high completion.[2,34,40–42] Average compensation per diary day was $1.18 (range: $0.14 – $3.00). The cell phone-based study, in addition to financial compensation, provided participants with free domestic phone calls, text messages, and phone-based Internet access during the study, with the option to retain the phone at study completion.[3]

Missing data can introduce bias into a study if appropriate analytical steps are not taken.[50] Among studies that reported how they treated missing data, three excluded participants with missing data,[3,40–44] three used modeling techniques that accounted for missing data,[29,34,35,49] and one used imputation.[33] Missing data are especially relevant for studies comparing diary and recall data; only one of the four studies that did so addressed data missingness, using imputation.[33]

Diary completion rates

No single measure of diary completion was reported by all studies (Table 2). Regardless of the measures provided, most studies reported relatively high completion. Nine reported that >80% of all diaries were submitted.[2,3,29,31–33,37,40–45,49] Four separately reported the proportion of diaries submitted on time; two reported on-time rates >90%.[32,47] To enable comparisons between studies, we divided the mean or median number of completed diaries by the number expected per subject to obtain a proportion. Four studies reporting means[3,30,40–42,47,48] and four reporting medians[2,3,30,34,48] had completion rates >80%. In a 12-month study of young MSM, Glick et al. noted a substantial decline in completion after six months,[33] however studies of a similar length in heterosexual populations did not report comparable declines.[29,31,45,49]

Table 2.

Diary Completion Rates among Sexual Risk Behavior Studies Using Web-Based Diaries

Study	Proportion of diaries submitted on time	Total proportion of diaries submitted	Proportion of subjects who reached specific completion level	Mean/median number of diaries completed per subject
Baer et al. (2002); Winer et al. (2006) [29,49]	---	90.7%	---	---
Wilson et al. (2008); Boone et al. (2013) [30,48]	---	78.0%	---	Mean: 5.2/6 Median: 6/6
Partridge et al. (2007); Edelstein et al. (2012) [31,45]	---	94.8%	---	---
Garry et al. (2002) [32]	99.8%	100.0%	94.6% completed all diaries on time	---
Glick et al. (2013) [33]	---	81.6%	Completed >80.0% of diaries: 82.0% in mo. 1–3; 72.0% in mo. 4–6; 57.1% in mo. 7–9; 55.6% in mo. 10–12	---
Grov et al. (2010) [34]	---	75.2%	---	Mean: 22.5/30 (SD=8.3) Median: 26/30
Hensel et al. (2012a) [3]	---	89.7%	82.3% completed 99.2% of diaries	Mean: 18.9/21 per week Median: 20/21 per week
Reece et al. (2008); Hensel et al (2011) [36,46]	---	61.9%	---	---
Hensel et al. (2012b) [35]	---	---	25.0% completed all diaries; 30.0% completed 50.1–99.9% of diaries	Mean: 13/30 (SD= 10)
Horvath et al. (2007) [37]	72.0%	99.7%	97.8% completed all diaries	---
Kiene et al. (2008); Kiene et al (2009) [38,39]	---	79.4%	---	Mean: 23.8/30 (SD=6.1)
Lim et al. (2010) [2]	62.6%	99.3%	98.5% completed at least one diary	Median: 13/13
Mustanski (2008); Mustanski (2007a); Mustanski (2007b) [40–42]	---	83.6%	70.0% completed at least 20/30	Mean: 24.4/30 (SD=9.9)
Newcomb (2013); Newcomb and Mustanski (2013) [43,44]	---	83.7%	---	---
Strachman and Impett (2009) [47]	93.8%	---	90.0% completed all diaries on time	Mean: 13/14 on time

We attempted to identify diary characteristics associated with high completion. For the studies in which appropriate data were available (n=9), we plotted the overall completion rate by compensation per diary day and found almost no association (R²=0.0047) (Figure 2, online supplement). We also compared completion rates by diary frequency and length of diary collection period, and found no significant differences.

Participant acceptability and privacy

Despite concerns about participant burden,[1] diaries were well-received. In three studies reporting acceptability, participants found electronic diaries to be convenient and enjoyed participating.[2,3,37] Two studies reported that participants felt the diaries sufficiently protected their privacy.[2,37]

Validity

Four studies used diaries to assess the validity of retrospective survey data and examined over- and underreporting in these surveys (Table 3). Garry et al. compared diary data to results from a surprise survey six to 12 months later.[32] On the survey, subjects underreported number of partners and overreported frequency of oral and vaginal sex and condom use. No difference was found for anal sex, the least frequent behavior.

Table 3.

Validity and Reactivity Findings for Studies Using Web-Based Diaries

Study	Findings
Validity ^
Garry et al. (2002) [32]	• Compared daily diaries to surprise retrospective survey 6–12 months later
	• Underreported number of partners (p=0.02) on survey
	• Overreported frequency of any sex, vaginal sex, and oral sex on survey (all p<0.01)
	• Overreported condom use for vaginal (p=0.02) and oral sex (p=0.06) on survey
	• Vaginal sex was reported 4 times more often in retrospective survey than diaries
	• No significant differences in recall of anal sex
	• Participants could not accurately predict their recall accuracy on retrospective survey
Glick et al. (2013) [33]	• Compared twice weekly, weekly, and biweekly diaries to 3 month retrospective survey
	• Non-significant underreporting on retrospective survey
	• Most CCC and Kappa statistics ≥80%
Horvath et al. (2007) [37]	• Compared daily diary to retrospective survey
	• Overreported receptive oral and anal sex on survey (all p<0.05)
	• Greater frequency of giving oral sex (p<0.05), receiving oral sex (p<0.01), and higher number of sexual events (p<0.01) were associated with overreporting on survey
	• Greater frequency of unprotected insertive anal sex was associated with underreporting on survey (p<0.001)
	• 28–85% of participants over or underreported each behavior on the retrospective survey
Lim et al. (2010) [2]	• Compared online, text-message, and paper weekly diaries to retrospective survey
	• Correlation was high for proportion of partners considered `regular' (0.87), frequency of sex (0.76), frequency of condom use (0.76), and STI risk (0.74)
	• Correlation was lower for number of partners (0.63), and proportion of partners considered `new' (0.57)
Reactivity
Glick et al. (2013) [33]	• Compared recall from active diary participants to controls
	• Controls had 0.9 more new male partners than diary subjects (p=0.05)
	• Controls had significantly greater increases in frequency of anal sex (p=0.01), frequency of unprotected anal sex (p<0.01), and any unprotected anal sex (p<0.01)
	• Controls had significantly higher STI and HIV diagnosis incidence rate (26.1%) than active diary participants (4.8%, p=0.01)
	• No evidence of dose-response by diary frequency
	• Author explanation: difference in HIV/STI incidence indicates that risky behavior may have actually differed over time between groups; attributed to diaries because of controlled nature of study
Hensel et al. (2012a) [3]	• Diary completion fell significantly over time (−0.61%/week, p<0.05)
	• Reports of vaginal sex decreased significantly over time (−0.61%/week, p<0.05)
	• Greater decrease in vaginal sex was reported in 18–20 year-olds (−0.91%/week), 27–29 year-olds (−0.83%/week), men (−0.80%/week), and patients without STI at enrollment (−0.68%/week) (all p<0.05)
	• Author explanation: slight differences in reported behavior during study may be due to subtle changes in behavior over time, not diary participation
Horvath et al. (2007) [37]	• Statistically significant decline in giving oral sex, receiving oral sex, insertive anal sex, and unprotected receptive anal sex over time (all p<0.05)
	• Author explanation: diary completion leads to self-monitoring, which causes participants to reduce risky behaviors over time
Kiene et al. (2008); Kiene et al (2009) [38,39]	• No evidence of reactivity
Newcomb (2013); Newcomb and Mustanski (2013) [43,44]	• No evidence of reactivity

^{^}Over- and underreporting on retrospective surveys compared to diaries

Horvath et al. found that MSM significantly overreported receptive oral and anal sex on a retrospective survey.[37] The proportion of participants who correctly reported number of sexual episodes and oral sex was low. Recall was most accurate for unprotected receptive anal sex. Higher diary reports of total sexual episodes, giving or receiving oral sex, and partner ejaculation during oral sex were associated with overreporting on the survey. Greater frequency of unprotected insertive anal sex was associated with underreporting.

Glick et al. assessed survey validity in a sample of MSM.[33] Concordance correlation coefficients and Kappa statistics exceeded 0.80 for almost all sexual behaviors assessed, indicating considerable agreement between methods.[33] Similarly, Lim et al. randomized young adults to complete weekly diaries online, on paper, or by text message, and compared results to a retrospective survey.[2] Correlation was highest for the proportion of regular vs. casual partners (0.87), mean frequency of sex (0.76), mean frequency of using condoms (0.76), and STI risk (0.74).[2]

Reactivity

Five studies assessed reactivity (Table 3), including one controlled study. Glick et al. compared quarterly surveys from participants randomized to an active diary or control group.[33] Compared to diary subjects, controls reported significantly greater increases over time in the occurrence of anal sex, frequency of anal and unprotected anal sex, and acquisition of new male partners. These behavioral differences may explain the significantly higher rate of incident HIV/STI diagnoses in control (26.1%) than diary subjects (4.8%, p=0.01).

Four studies assessed reactivity by analyzing intrapersonal temporal trends in behavior, without comparing to a control. In their cell phone study, Hensel et al. noted that diary submission rates declined significantly each week, demonstrating completion reactivity, and reports of vaginal sex declined significantly over time, demonstrating behavioral reactivity.[3] Horvath et al. examined trends in behavior in a diary study of MSM and found that giving and receiving oral sex, insertive anal sex, and receptive unprotected anal sex decreased significantly over time.[37] Two studies concluded that their data showed no evidence of reactivity.[38,39,43,44]

DISCUSSION

Web-based diaries are increasingly popular for collecting detailed sexual behavior data. To date, most studies have been small and not designed to assess methodological questions. There does not appear to be one single web-diary design that best measures sexual risk behavior, and individual study needs should guide methodological decisions regarding diary frequency, medium, and content. Nevertheless, our review identified interesting patterns and lessons learned across studies that implemented web-based diaries.

Diary completion rates in the reviewed studies were generally high. Nearly all studies provided compensation, but the amount of compensation was not associated with completion. While most collected data were based on daily behavior, researchers succeeded using a variety of diary submission schedules. In one study that directly compared three submission schedules, the less frequent schedules had higher completion rates.[33] Most diary studies lasted for one month, although some had success with more frequent (e.g., thrice daily) or longer duration data collection. In a recent study of young MSM, the decline in completion rates over time may have resulted from higher frequency of sexual behavior relative to other populations. This suggests that future studies should employ strategies to simplify data collection and encourage long-term participation. Inconsistent reporting of completion data and diary measures hindered our ability to further identify methods associated with improved completion. To enhance comparability, authors should at minimum report the total proportion of diaries received and should consider following the guidelines for reporting completion rates proposed by Stone and Shiffman,[51] including reporting what constituted complete and on-time submissions.

Several studies used diaries to assess the validity of retrospective surveys. Condom use, partner type, frequency, and type of sex were almost universally measured, both as partner-specific and aggregate variables. Overall, correlation between diary and retrospective data was high, indicating that both methods likely provide valid estimates. However, accuracy tended to be greater for less frequent behaviors, and overreporting on retrospective surveys exceeded underreporting. Clearly, study aims and data needs should be considered when selecting a data collection tool. This appears to be especially true for high-frequency behaviors (e.g., number of sex acts) where diary data likely provide more accurate count data.

It is difficult to draw firm conclusions about reactivity in web-based sexual risk behavior diaries given the paucity of quality studies. The notion that “self-monitoring” can generate behavior change has been studied within the context of many health behaviors – notably in exercise and weight loss intervention research – although the findings of controlled studies have overall been equivocal. Nevertheless, among those studies evaluating the effect of diaries (both web-based and non-web-based) on sexual behaviors, diary completion has been consistently associated with lower rates of sexual risk behavior.[33,52–54] Although further controlled research with more diverse populations is needed to confirm these findings, available data suggest that researchers using diaries should consider that this methodology may, in fact, be an intervention itself.

This review has two main limitations. First, we could not assess the presence of publication bias, but it is possible that web-based diary studies with non-significant findings or inadequate completion were less likely to be published. Second, the broad parameters used to identify articles minimized the risk of missing eligible records, but it is still possible that our strategy did not identify all relevant articles.

Our review highlights several areas for future research on web-based diaries specific to sexual risk behavior. Research should identify subjects' preferred diary mediums, frequency, and length, and use diaries with populations beyond MSM and university students. Cell phone diaries possess largely unexplored potential, as apps for smartphones can collect data, transmit to study staff, and enable staff to remind participants to complete diaries. This medium is particularly compelling for research and prevention efforts among traditionally hard-to-reach populations, as demonstrated in one study of low-income STI clinic attendees.[3] In addition, phone-based diaries may be especially effective among young black MSM – a priority population for HIV/STI prevention research in the U.S. – given that smartphone use appears to be extremely common and accepted as an intervention tool.[55] Studies should also examine methods, including compensation, that maximize completion. Reactivity studies should incorporate control groups to avoid conflating temporal trends with behavioral reactivity. These studies could also identify features that affect reactivity, such as frequency, length of collection, or medium. Long-term studies should examine if reactivity is temporary or long-lasting. Finally, there is limited information on how missing diary data are handled analytically. Future research should examine if diary entries are missing at random, which would inform the use of imputation.

Web-based diaries provide many benefits, including remote data collection and reduced data entry burden. However, the potentially substantial costs of software, hardware, and privacy requirements associated with implementing a web-based methodology indicate that researchers should consider their data needs and population when selecting a web-based diary.[1,12,56,57] This tool is a promising technological advancement for HIV/STI research, and future studies should continue to employ rigorous study designs to clarify the most appropriate methodological practices.

KEY MESSAGES.

• To date, most web-based, quantitative, sexual risk behavior diaries have been daily, month-long studies collected using websites.

• Methods that maximize diary completion are unclear, partially due to non-systematic reporting of completion rates.

• Compared to diaries, retrospective surveys appear to collect valid measures of sexual risk behavior, but may overestimate the frequency of more common behaviors.

• Reactivity due to diary completion may decrease sexual risk behavior, but additional controlled studies are needed.