Prospective Memory in HIV Infection: Is “Remembering to Remember” a Unique Predictor of Self-reported Medication Management?

Abstract

Optimal adherence to antiretroviral medications is critical to the effective long-term management of HIV infection. Although prospective memory (ProM; i.e., “remembering to remember”) has long been theorized to play an important role in medication adherence, no prior studies have evaluated whether HIV-associated ProM impairment possesses unique predictive value in this regard. Results from this study demonstrate a robust association between ProM impairment and self-reported medication management in 87 HIV-infected persons currently prescribed antiretroviral medications. Specifically, more frequent ProM complaints and performance deficits on both laboratory and semi-naturalistic ProM tasks were all independently related to poorer self-reported medication management. A series of hierarchical regression analyses revealed that HIV-associated ProM impairment accounted for a significant amount of variance in self-reported medication management beyond that which was explained by other factors known to predict nonadherence, including mood disorders, psychosocial variables, environmental structure, and deficits on a traditional battery of neuropsychological tests. Overall, these findings support the hypothesis that ProM captures a unique and largely untapped aspect of cognition that is germane to optimal medication adherence. The potential benefits of individualized remediation strategies that are informed by conceptual models of ProM and specifically target medication adherence warrant further exploration.

INTRODUCTION

The advent and widespread use of effective combined antiretroviral therapies (cART) has resulted in a dramatic decrease in mortality for individuals living with the human immunodeficiency virus (HIV) in the developed world (CDC, 2005). Strict adherence to prescribed medication regimens (i.e., cART) is essential to the long-term clinical management of HIV disease. Infected individuals who are highly adherent (i.e., taking over 90% of their prescribed doses) evidence considerably better disease outcomes, including lower rates of virologic failures (e.g., Perno et al., 2002), treatment resistant viral strains (e.g., Harrigan et al., 2005), and mortality (e.g., Lima et al., 2007). Identifying salient risk factors for nonadherence is therefore of considerable public health importance, both in terms of facilitating early detection and informing targeted cognitive and behavioral interventions to maximize adherence. Conceptual models of adherence in HIV highlight the numerous complexities of this health behavior (e.g., Barclay et al., 2007; Rosenstock, 1974; Starace et al., 2002), including the contributions of psychiatric symptoms (e.g., depression), health beliefs (e.g., attitudes toward medicine), social variables (e.g., social support), demographics (e.g., ethnicity), and environmental factors (e.g., use of compensatory strategies).

Neuropsychological impairment is another important predictor of nonadherence in HIV (e.g., Hinkin et al., 2002). Per patient self-report, “forgetting” is among the most frequently perceived causes of missing antiretroviral (ARV) doses (Chesney et al., 2000). Indeed, objective deficits in episodic memory, as well as in executive functions, are associated with increased risk of ARV nonadherence (e.g., Hinkin et al., 2002; 2004; cf. Waldrop-Valverde et al., 2006), even after considering the effects of psychiatric and demographic factors. Further underscoring the importance of cognitive risk factors, mild-to-moderate deficits in numerous areas of neuropsychological functioning (e.g., memory, executive functions, information processing speed) remain prevalent in the era of cART, despite the effectiveness of such treatments on immune health. As such, a vicious cycle may be kindled whereby nonadherent cognitively impaired individuals experience more rapidly advancing HIV disease, thereby potentially accelerating HIV-associated neural injury and neuropsychological impairment, which in turn reduces the likelihood of optimal adherence (Albert et al., 1999).

One cognitive construct with obvious relevance to medication adherence is prospective memory (ProM). ProM, or “remembering to remember,” is an aspect of declarative (i.e. episodic) memory that refers to the execution of a future intention (e.g., remembering to pay one’s rent at the beginning of the month). Although conceptual models vary somewhat across the literature (see McDaniel & Einstein, 2007), ProM generally involves a series of cognitive processes that includes: 1) the formation of an intention that is paired with a specific retrieval cue, (i.e., a specific time or event); 2) maintenance of the intention-cue pairing over a delay interval while concurrently engaged in a foreground task, during which time both active (e.g., strategic) and automatic (i.e., spontaneous) monitoring may occur (McDaniel & Einstein, 2007); 3) detection and recognition of the cue; 4) search and retrieval from retrospective memory (RetM) for the content of the intention; and 5) successful execution of the intention. At a neural systems level, ProM is most heavily reliant upon prefrontal systems (e.g., Brodmann’s area 10; Simons et al., 2006), but is also dependent upon the RetM contribution of medial temporal (e.g., hippocampal) networks (e.g., Martin et al., 2007^b). From a more practical perspective, ProM is hypothesized to play a unique and influential role in numerous aspects of daily life, including household chores (e.g., cooking; Fortin et al., 2002) and employment (e.g., Sellen et al., 2007).

It is widely believed that ProM impairment carries the strongest day-to-day effects on medication adherence (Park & Kidder, 1996). In fact, most publications on ProM highlight the clinical relevance of this construct by referencing its seemingly obvious implications for successful medication management. At a conceptual level, normal ProM is critical in order to properly adhere to a medication regimen; that is, one must first form the intention to take the medication at some point in the future (e.g., take two pills of medication X after dinner), maintain the intention-cue pairing throughout the day despite the distraction of normal daily activities, detect and recognize the cue when it occurs (e.g., clearing the dinner table), recall the specific medication and directions, and finally, take the medication as instructed. In fact, it has been argued that ProM may be more critical to adherence than other cognitive ability areas (Woods et al., in press^a); for example, using the scenario above, the complex process of remembering to take the medication as prescribed (i.e., ProM) is arguably more critical than is recalling the medication’s name upon being queried (i.e., RetM).

Surprisingly, however, our review of the literature yielded only two peer-reviewed articles that have specifically examined the relationship between ProM and medication adherence. In a study of older adults with Type 2 Diabetes, Vedhara et al. (2004) demonstrated that participants who failed a habitual ProM task were significantly less adherent to their oral hypoglycemic medications. Hertzog and colleagues (2000) found a modest relationship between ProM complaints and medication adherence in individuals with rheumatoid arthritis; however, adherence was not associated with performance on a semi-naturalistic ProM task (i.e., making scheduled daily telephone calls to the examiner). Thus, despite its conceptual attractiveness, it remains largely unknown whether ProM is an important predictor of medication nonadherence in HIV or if it possesses any incremental ecological validity relative to established cognitive (e.g., RetM and executive functions) and non-cognitive (e.g., affective distress and psychosocial factors) predictors of suboptimal adherence.

This study aimed to elucidate the role of ProM in medication management among persons living with HIV infection. Recent research shows that individuals infected with HIV report experiencing more frequent ProM failures as compared to demographically similar seronegative persons, especially on self-cued daily tasks (Woods et al., 2007^a). Mild-to-moderate impairment is also evident on performance-based measures of both time- and event-based ProM (e.g., Carey et al., 2006; Martin et al., 2007^a). Consistent with the largely frontal systems neuropathophysiology of HIV (e.g., Wiley et al., 1998), the profile of HIV-associated ProM impairment is characterized by deficient self-initiated cue detection and retrieval (Carey et al., 2006). Most importantly, ProM demonstrates incremental ecological validity as a predictor of HIV-associated neurocognitive disorders; that is, ProM deficits account for significant variance in general decline in instrumental activities of daily living (IADL) beyond that which is explained by RetM impairment and affective distress (Woods et al., in press^a). Therefore, it is hypothesized that: 1) HIV-associated ProM impairment is associated with poorer medication adherence; 2) HIV-associated ProM impairment is predictive of poorer medication adherence, even after considering the contributions of deficits in other cognitive ability areas (e.g., RetM); and 3) HIV-associated ProM impairment is a unique predictor of medication adherence relative to established non-cognitive predictors of adherence (e.g., psychiatric comorbidity, demographics, and psychosocial variables).

METHOD

Participants

The study sample was comprised of 87 participants with HIV infection who were currently prescribed at least one ARV. Table 1 displays the demographic, HIV disease, and psychiatric characteristics of the cohort, which was recruited from the general community (e.g., newspaper advertisements) and local HIV clinics (e.g., posted flyers). Exclusion criteria included severe psychiatric illness (e.g., schizophrenia or bipolar disorder), neurological disease (e.g., seizure disorders, stroke, closed head injuries with loss of consciousness greater than 15 minutes, and central nervous system neoplasms or opportunistic infections), verbal IQ < 70 (based on the Wechsler Test of Adult Reading; Psychological Corporation, 2001), substance dependence within six months of evaluation, and a urine toxicology screen positive for illicit drugs (other than marijuana) on the day of evaluation. Individuals with positive marijuana toxicology screens (n = 12) were not excluded because marijuana is detectable up to 30 days after last use and medications commonly prescribed for persons with HIV (e.g., efavirenz, marinol) can produce positive toxicology results.

Table 1.

Participants’ demographic, disease, and psychiatric characteristics

Variable	Descriptive statistics (N = 87)
Demographic Characteristics
Age (years)	45.9 (8.8)
Education (years)	13.8 (2.8)
Estimated Verbal IQa	105.8 (13.0)
Sex (% women)	10.3
Ethnicity (% Caucasian)	70.1
HIV Disease Characteristics
Duration of infection (years)	14.1 (7.0)
Current CD4 count (cells/μl)	567.5 (315.6)
Plasma HIV RNA (log10)	1.9 (0.8)
AIDS diagnoses (%)	59.8
cART (%)	98.9
Hepatitis C co-infection (%)b	12.2
Duration of current ARV regimen (months)	26.0 (33.4)
Pill burden (total no. prescribed medications)	7.3 (4.2)
Total no. ARVs	3.9 (1.0)
Total no. nonARVs	3.3 (3.9)
Psychiatric Characteristics
Lifetime substance dependence (%)c	55.3
Alcohol	31.0
Cocaine	21.2
Marijuana	11.5
Methamphetamine	27.6
Major depressive disorder (%)
Currentd	8.2
Lifetime	45.9
Generalized anxiety disorder (%)
Currentd	2.4
Lifetime	7.1

Note. Data are presented as means and standard deviations, unless otherwise indicated. CD4 = cluster of differentiation 4; cART = combination antiretroviral therapy; ARV = antiretroviral.

^aVerbal IQ was derived from the Wechsler Test of Adult Reading (WTAR; M = 100, SD = 15).

^bN = 82.

^cDenotes a lifetime DSM-IV diagnosis of dependence on any substance.

^dRefers to the proportion of individuals who met DSM-IV criteria within the month prior to evaluation.

Materials and Procedure

All participants provided written, informed consent prior to completing a comprehensive medical, psychiatric, and neuropsychological research evaluation, which was approved by the institution’s human research protections program. The primary dependent variable of interest was self-reported medication management, which was assessed with the 20-item “Medication Management Efficacy Scale (MMES)” (e.g., “I am less efficient at adhering to my medication regimen than I used to be”) of the Beliefs Related to Medication Adherence (BERMA; McDonald-Miszczak et al., 2004) questionnaire. All BERMA MMES items were rated on a 5-point Likert scale, ranging from 1 (strongly disagree) to 5 (strongly agree) (NB. several items require reverse scoring so that higher values reflect better perceived medication management efficacy). Prior data support the reliability and factor structure of the BERMA (McDonald-Miszczak et al., 2004). In the current study sample (N = 87), the split-half reliability (i.e., Spearman-Brown) of the BERMA MMES scale was .91 and its internal consistency (i.e., Cronbach’s alpha) was .93.

Prospective Memory Assessment

Participants completed the Memory for Intentions Screening Test (MIST; Raskin, 2004), which is a standardized laboratory measure of ProM with published evidence of its reliability (Woods et al., in press^b) and construct validity (Carey et al., 2006; Woods et al., 2006; 2007^a; in press^a,^b). The MIST is a 30-min task in which eight different intentions are prescribed while the participant completes a word search puzzle (see Woods et al., 2007^b; in press^b). The following variables were derived from the MIST: 1) summary score; 2) time-based scale; 3) event-based scale; 4) distractor total; 5) recognition total; and 6) a retrieval index (see Carey et al., 2006). In addition, the following error types were coded: 1) no response (i.e., cue omissions); 2) task substitutions (e.g., substitution of an action response for a verbal task or vice-versa); 3) loss of content (e.g., examinee acknowledgment that a response is required to a cue, but nevertheless fails to recollect its content); and (4) loss of time (i.e., executing a correct intention that is ± 15% of the target time). Finally, a 24-hour probe is given, instructing participants to leave a telephone message for the examiner the following day, specifying the number of hours slept the night after the assessment (score range = 0–2). The 24 hr delay is an adjunct trial to the MIST and is not factored into the summary score. Unlike the other MIST items, participants are allowed to use mnemonic strategies for this trial (e.g., a note in their electronic organizer), but are not explicitly prompted to do so.

Self-reported ProM was assessed using the ProM Scale from the Prospective and Retrospective Memory Questionnaire (PRMQ; Smith et al., 2000), which consists of eight ProM-specific items that are rated on a 5-point Likert-type scale that ranges from 1 (“never”) to 5 (“very often”). The PRMQ ProM complaints are separated into four self-cued (e.g., “How often do you forget appointments if you are not prompted by someone else or by a reminder, such as a diary or a calendar?”) and four environmentally-cued (e.g., “How often do you forget to buy something you planned to buy, like a birthday card, even when you see the shop?”) ProM complaints. Total scores on the PRMQ ProM Scale can range from 8 to 40.

Standard Neuropsychological Assessment

All participants were administered a standardized battery of clinical neuropsychological tests, which were selected in accordance with NIH guidelines to cover seven cognitive domains commonly impaired in HIV disease (Butters et al., 1990). Specifically, the information processing speed domain included Trail Making Test, Part A (TMT; Reitan & Wolfson, 1985) and total execution time from the Tower of London – Drexel (ToL-DX; Culbertson & Zillmer, 2001). Motor skills were assessed using the dominant and nondominant hand total scores from the Grooved Pegboard test (Kløve, 1963). Measures of executive functions included TMT Part B and the ToL-DX total move score. Tests in the attention domain included the Digit Span subtest from the Wechsler Adult Intelligence Scale – Third Edition (The Psychological Corporation, 1997) and Trial 1 from the California Verbal Learning Test – Second edition (CVLT-II; Delis et al., 2000). The retrospective learning and memory domains were comprised of the immediate and long delayed trials of the CVLT-II and the Rey-Osterreith Complex Figure (Stern et al., 1999). Finally, verbal fluency was assessed with the animal (Benton et al., 1994) and action (Woods et al., 2005) fluency tests. For each of the seven cognitive domains, blind clinical ratings of impairment were generated from published, demographically-adjusted T-scores by a neuropsychologist using standardized methods (Heaton et al., 1994; Woods et al., 2004). In brief, clinical ratings were based on a scale that ranges from one (above average, T-score ≥ 55) to nine (severely impaired, T-score < 20), where scores of five and higher (T-score < 40) represent cognitive impairment per the guidelines recommended by Heaton and colleagues (Heaton, Miller, Taylor, & Grant, 2004). Slightly over 32% (n = 28) of the sample demonstrated global neuropsychological impairment.

Psychiatric Assessment

All participants underwent a structured psychiatric assessment using the Composite International Diagnostic Interview (CIDI version 2.1; World Health Organization, 1998) to generate lifetime and current (i.e., within 1 month of evaluation) diagnoses of Major Depressive Disorder (MDD), Generalized Anxiety Disorder (GAD), and Substance-Related Disorders per Diagnostic and Statistical Manual of Mental Disorders (4^th ed.; American Psychiatric Association, 1994) criteria. Acute affective distress was measured with the Profile of Mood States (POMS; McNair et al., 1981). The POMS is a 65-item, self-report measure of current (i.e., the week prior to evaluation) mood states in which participants rate various adjectives (e.g., “sluggish”) on a five-point Likert-type scale ranging from 0 (“not at all”) to 4 (“extremely”). Items on the POMS comprise six subscales (i.e., Confusion, Depression, Fatigue, Anger, Vigor, and Tension) and a Total Mood score, for which higher scores indicate greater distress.

Psychosocial and Environmental Factors

Two psychosocial scales of relevance to medication management were derived from the BERMA. The “Dealing with Health Professionals” scale contains 23 items that measure the strength of participants’ communication and relationship with their medical providers (e.g., “I have difficulty talking openly with my physician”). The “Attitudes About Medications” scale is comprised of 10 items measuring participants’ health beliefs (e.g., “I am taking too much medication for my medical conditions”). Both subscales are rated on the same Likert scale as the BERMA MMES described above.

The Prospective Memory for Medications Questionnaire (PMMQ; Gould, McDonald-Miszczak, & Gregory, 1997) is a self-report questionnaire with 33 questions assessing how often an individual uses different cognitive (e.g., “Do you regularly repeat to yourself the instructions for taking a prescription…?”) and behavioral (e.g., “Do you use a clock or watch alarm to remind you when it is time to take your medications?”) medication taking strategies. Participants are asked to rate how often they use each strategy on a 5-point Likert-type scale, ranging from 0 (“never”) to 4 (“always”), whereby higher scores indicate more frequent strategy use.

Medical Assessment

Finally, participants also received a full neuromedical evaluation, which included a thorough review of medications, medical history and current symptoms, a complete physical and neurological evaluation, CDC staging, and a blood draw. HIV serostatus was determined by enzyme linked immunosorbent assays and confirmed by a Western Blot test. Standard flow cytometry methods were used to count CD4+ lymphocytes in blood samples. Plasma HIV RNA levels were quantified using RT-PCR (Amplicor, Roche Diagnostics, Indianapolis, IN).

Data Analyses

A series of correlations and a follow-up linear regression were used to examine the associations between the BERMA MMES and measures of ProM. In order to consider the contribution of other important predictors of medication management, we conducted a series of correlations (or between-group significance tests and Cohen’s d effect size estimates for categorical variables) with the BERMA MMES and representative variables selected on an a priori basis from the following domains: 1) general neuropsychological functioning; 2) psychiatric status; 3) psychosocial and environmental factors; and 4) demographic characteristics. In order to determine whether objective ProM was a unique predictor of self-reported medication management (i.e., the BERMA MMES), we then conducted separate hierarchical regression models for each domain. In the first step of these models, we entered only those predictor variables that demonstrated significant relationships with the BERMA MMES in the preliminary analyses. The second step of each model included the performance-based (i.e., MIST Summary Score), self report (i.e., PRMQ) and semi-naturalistic (i.e., 24 hr delay trial) indicators of ProM. A critical alpha level of .05 was used for all analyses.

RESULTS

Descriptive data for the entire sample (N = 87) on the BERMA, MIST, PRMQ and the seven cognitive domains are presented in Table 2.

Table 2.

Descriptive data for the BERMA, MIST, PRMQ, and standard cognitive tests (N = 87)

Variable	Mean	SD	Median	IQR	Range
BERMA
Medication management	77.6	13.2	77.0	71.0, 88.0	42.0, 100.0
Dealing with health professionals	94.2	14.0	95.0	84.0, 106.0	64.0, 115.0
Attitudes about medications	36.5	6.3	36.0	32.0, 42.0	23.0, 49.0
MIST
Summary score	38.8	7.3	39.0	36.0, 45.0	15.0, 48.0
Time-based	5.9	1.4	6.0	5.0, 7.0	2.0, 8.0
Event-based	7.0	1.5	7.0	7.0, 8.0	0.0, 8.0
Error types
No response	0.7	1.0	0.0	0.0, 1.0	0.0, 5.0
Task substitutions	0.6	0.8	0.0	0.0, 1.0	0.0, 5.0
Loss of content	0.7	0.8	0.0	0.0, 1.0	0.0, 3.0
Loss of time	0.4	0.6	0.0	0.0, 1.0	0.0, 3.0
Recognition post-test	7.8	0.5	8.0	8.0, 8.0	6.0, 8.0
Retrieval index	2.2	1.5	2.0	1.0, 3.0	0.0, 6.0
Word search distracter	16.5	5.6	16.0	13.0, 19.0	6.0, 37.0
24 hr delay	1.0	0.8	1.0	0.0, 2.0	0.0, 2.0
PRMQ
ProM total score	19.8	5.8	19.0	16.0, 24.0	8.0, 39.0
Self-cued scale	10.3	3.0	10.0	8.0, 12.0	4.0, 19.0
Environmentally-cued scale	9.5	3.0	9.0	7.0, 11.0	4.0, 20.0
Other Cognitive Domains
Global cognitive functioning	3.9	1.4	4.0	3.0, 5.0	1.0, 8.0
Information processing speed	2.6	1.3	2.0	2.0, 4.0	1.0, 6.0
Retrospective learning	2.6	1.5	2.0	2.0, 4.0	1.0, 7.0
Retrospective memory	2.6	1.5	2.0	2.0, 3.0	1.0, 8.0
Executive functions	2.7	1.6	2.0	2.0, 4.0	1.0, 7.0
Attention	3.0	1.4	3.0	2.0, 4.0	1.0, 8.0
Verbal fluency	3.0	1.4	3.0	2.0, 4.0	1.0, 6.0
Motor coordination	2.6	1.6	2.0	2.0, 3.0	1.0, 9.0

Note. BERMA = Beliefs Related to Medication Adherence; IQR = interquartile range; MIST = Memory for Intentions Screening Test; PRMQ = Prospective and Retrospective Memory Questionnaire; ProM = prospective memory; SD = standard deviation.

Table 3 displays the correlations between the BERMA MMES and the various ProM measures. The BERMA MMES was positively correlated with the MIST summary score (p < .05), such that better adherence was associated with higher ProM performance. Analysis of the various MIST components revealed the strongest relationships between the BERMA MMES and the time- and event-based scales, no response errors, word-search distracter task, retrieval index, and 24 hr delay trial (ps<.05). Due to concerns regarding colinearity (i.e., several MIST interscale correlations > .7), two separate follow-up regression models were conducted based on conceptual groupings of these significant variables as predictors of the BERMA MMES. The first regression model, which included only the primary indices (i.e., PRMQ total score, the MIST time- and event-based subscales, and the 24hr delay) was significant, R²=.37, F(4,82)=13.4, p<.0001. Significant, unique contributors to this model included the PRMQ, MIST time-based scale, and 24hr delay (ps<.05). The MIST distractor score and retrieval index were entered into the second regression model, which was also significant, R²=.10, F(2,82)=5.7, p=.005, as driven exclusively by the retrieval index (p=.005).

Table 3.

Correlations between the BERMA MMES and the ProM measures (N = 87)

Cognitive Variable	Correlation Coefficienta	p
MIST
Summary score	0.36	.001
Time-based	0.34	.001
Event-based	0.30	.005
Error types
No response	−0.37	.0004
Task substitutions	−0.17	.121
Loss of content	−0.06	.554
Loss of time	−.02	.854
Recognition post-test	0.13	.247
Retrieval index	−0.30	.006
Word search distracter	0.28	.008
24 hr delay	0.33	.002
PRMQ
PRMQ ProM total score	−0.47	<.0001
ProM self-cued scale	−0.44	<.0001
ProM environmentally-cued scale	−0.47	<.0001

Note. BERMA MMES = Beliefs Related to Medication Adherence Medication Management Efficacy Scale; MIST = Memory for Intentions Screening Test; PRMQ = Prospective and Retrospective Memory Questionnaire; ProM = prospective memory.

^aSpearman’s rho.

As shown in Table 4, the general cognitive correlates of the BERMA MMES included the domains of executive functions, attention, and verbal fluency (ps<.05), along with a trend-level finding for RetM learning (p<.10). A hierarchical regression model that included all of the significant traditional cognitive domains in the first step (see Table 4) predicted the BERMA MMES at a trend level (p<.10). However, the addition of the ProM variables (i.e., MIST summary score, MIST 24 hr trial, and the PRMQ) in the second step of the model resulted in a significant increase in the amount of variance in medication management explained by the traditional cognitive factors (p<.001).

Table 4.

Cognitive predictors of the BERMA MMES (N = 87)

Variable	Statistics
Correlational analyses	ρa		p
Global cognitive functioning	−0.17		.112
Information processing speed	−0.04		.700
Retrospective learning	−0.21		.056
Retrospective memory	−0.12		.227
Executive functions	−0.26		.014
Attention	−0.22		.038
Verbal fluency	−0.28		.008
Motor coordination	−0.03		.767
Heirarchical regression	B	SEB	β	R2	ΔR2
Step 1				0.06†
Retrospective learning	−0.31	1.03	−0.04
Executive functions	−0.41	0.92	−0.05
Attention	−1.36	1.11	−0.15
Verbal fluency	−2.36	1.06	−0.25*
Step 2				0.35***	0.30***
MIST summary score	0.38	0.17	0.21*
24 hr trial	3.31	1.46	0.21*
PRMQ ProM scale	−1.08	0.21	−0.47***

Note. BERMA MMES = Beliefs Related to Medication Adherence Medication Management Efficacy Scale; MIST = Memory for Intentions Screening Test; PRMQ = Prospective and Retrospective Memory Questionnaire.

^aSpearman’s rho.

^†p < .10.

^*p < .05.

^***p < .001.

Current affective distress as measured by the POMS Total Mood score and its six subscales was also significantly correlated with the BERMA MMES (ps<.01; see Table 5), such that greater levels of affective distress were associated with poorer self-reported medication management. Interestingly, a lifetime diagnosis of GAD was associated with better medication management (p<.05), whereas a diagnosis of current MDD was associated with poorer medication management (p<.05). No other psychiatric diagnosis was independently associated with the BERMA MMES, including current GAD, lifetime MDD, any lifetime substance-related disorder, or a positive urine toxicology screen for marijuana (ps<.10). When lifetime GAD, current MDD, and the POMS Total Mood score were entered in the first step of a regression predicting the BERMA MMES (see Table 5), the model was significant (p<.001). In the second step, the MIST summary score, MIST 24 hr trial, and PRMQ significantly increased the proportion of variance explained by psychiatric factors alone (p<.001).

Table 5.

Psychiatric predictors of the BERMA MMES (N = 87)

Variable	Statistics
Basic analyses	Effect size		p
Profile of Mood States total	−0.52a		<.0001
Tension	−0.45a		<.0001
Depression	−0.48a		<.0001
Anger	−0.48a		<.0001
Vigor	0.29a		.007
Fatigue	−0.38a		.0002
Confusion	−0.56a		<.0001
Major Depressive Disorder
Current	0.98b		.024
Lifetime	−0.34b		.200
Generalized Anxiety Disorder
Current	−1.45b		.101
Lifetime	−1.16b		.029
Substance dependence (lifetime)	0.24b		.168
Heirarchical regression	B	SEB	β	R2	ΔR2
Step 1				0.25***
POMS total	−0.21	0.05	−0.54***
MDD current	9.32	4.93	0.18†
GAD lifetime	5.18	5.96	0.11
Step 2				0.39***	0.16***
MIST summary score	0.33	0.17	0.18*
24 hr trial	2.67	1.41	0.17†
PRMQ ProM scale	−0.85	0.23	−0.36***

Note. BERMA MMES = Beliefs Related to Medication Adherence Medication Management Efficacy Scale; MIST = Memory for Intentions Screening Test; PRMQ = Prospective and Retrospective Memory Questionnaire; POMS = Profile of Mood States; MDD = major depressive disorder; GAD = generalized anxiety disorder.

^aSpearman’s rho.

^bCohen’s d.

^†p < .10.

^*p < .05.

^***p < .001.

The psychosocial and environmental predictors of the BERMA MMES included the other two BERMA scales (i.e., Dealing With Health Professionals and Attitudes About Medications), and the behavioral scale of the PMMQ (all ps<.001). As shown in Table 6, regression including these factors as predictors of the BERMA MMES was significant (p<.0001). Once again, the inclusion of the MIST summary score (p<.001) improved the overall model (p=.001), whereas PRMQ contributed only at a trend level (p<.10) and the MIST 24 hr trial was not significant (p>.10).

Table 6.

Psychosocial and environmental predictors of the BERMA MMES (N = 87)

Variable	Statistics
Correlational analyses	ρa		p
BERMA
Dealing with health professionals	0.85		<.0001
Attitudes about medications	0.65		<.0001
PMMQ Total	−0.27		.013
Cognitive strategies	−0.17		.132
Behavioral strategies	−0.39		.0002
Heirarchical regression	B	SEB	β	R2	ΔR2
Step 1				0.69***
BERMA health professionals	0.66	0.08	0.70***
BERMA attitudes	0.20	0.18	0.09
PMMQ total	−0.32	0.14	−0.15*
Step 2				0.74***	0.06**
MIST summary score	0.35	0.11	0.19**
24 hr trial	1.15	1.03	0.07
PRMQ ProM scale	−0.29	0.15	−0.13†

Note. BERMA MMES = Beliefs Related to Medication Adherence Medication Management Efficacy Scale; MIST = Memory for Intentions Screening Test; PRMQ = Prospective and Retrospective Memory Questionnaire; PMMQ = Prospective Memory for Medications Questionnaire.

^aSpearman’s rho.

^†p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

Concerning HIV disease and treatment characteristics, the BERMA MMES was unrelated to duration of infection, current and nadir CD4 lymphocyte count, plasma HIV RNA, AIDS status, co-infection with hepatitis C, duration of ARV regimen, or pill burden (all ps>.10). Similarly, the BERMA MMES was not significantly associated with age, education, or sex (ps>.10); however, Caucasian participants reported better perceived adherence as compared to individuals of other ethnic identities (p<.05). A hierarchical regression model predicting the BERMA MMES from ethnicity was significant (p<.05; see Table 7), with significant additional variance explained by including the MIST summary score, MIST 24 hr trial, and PRMQ variables (p<.001).

Table 7.

Demographic predictors of the BERMA MMES (N = 87)

Variable	Statistics
Basic analyses	Effect size		p
Demographics
Age	−0.10a		.35
Education	0.11a		.33
Sex	−0.26b		.336
Ethnicity	−0.50b		.017
Heirarchical regression	B	SEB	β	R2	ΔR2
Step 1				0.04*
Ethnicity	6.48	3.04	.23*
Step 2				0.39***	0.37***
MIST summary score	0.40	0.16	0.22*
24 hr trial	2.76	1.39	0.17*
PRMQ ProM scale	−1.11	0.20	−0.49***

Note. BERMA MMES = Beliefs Related to Medication Adherence Medication Management Efficacy Scale; MIST = Memory for Intentions Screening Test; PRMQ = Prospective and Retrospective Memory Questionnaire.

^aSpearman’s rho.

^bCohen’s d.

^*p < .05.

^***p < .001.

DISCUSSION

It has long been hypothesized that ProM plays a critical role in medication adherence; however, very few empirical studies have examined this assertion. To our knowledge, this is the first comprehensive study of ProM as a predictor of medication management in HIV infection. Results revealed that ProM was robust and valuable predictor of self-reported medication management in 87 HIV-infected persons currently taking ARVs, which is commensurate with a prior study in older adults with diabetes (Vedhara et al., 2004). Specifically, more frequent ProM complaints and performance deficits on both laboratory and semi-naturalistic ProM tasks were all related to poorer medication management, over and beyond other valid cognitive (e.g., executive functions), psychiatric, psychosocial, and environmental predictors of adherence.

At the level of component cognitive processes, the strongest relationships between HIV-associated ProM impairment and medication management were observed for time-based trials and omission (i.e., no response) errors. From a conceptual perspective, these data suggest that successful medication management may depend heavily upon monitoring, cue detection, and perhaps intention retrieval, rather than encoding or consolidation (as might be suggested by loss of content errors and impaired recognition). These findings indicate that poorer medication management in HIV is associated with impairment in the strategic allocation of cognitive resources that are required to effectively navigate the concurrent demands of performing an ongoing task while simultaneously monitoring and detecting cues. Indeed, these results converge with data showing that “forgetting” and “being busy” are two of the most common reasons that HIV-infected individuals miss medications doses (Chesney et al., 2000). Overt strategic monitoring is particularly important over brief intervals when retrieval cues are not central to the ongoing activities (e.g., McDaniel & Einstein, 2007), as is the case with time-based ProM tasks used in this study. In other words, independent tracking of time requires more overt strategic monitoring than simply recognizing associative cues that are embedded in the foreground task.

Automatic ProM monitoring processes may also contribute to adherence as they are hypothesized to play a stronger role in ProM over longer delay intervals (e.g., days or weeks). To this end, performance on the 24 hour ProM delay of the MIST was predictive of medication management, independent of the performance-based and self-report measures of ProM administered in the laboratory. Prior research shows that individuals with HIV are over three times more likely to fail this particular task relative to demographically comparable seronegatives (Carey et al., 2006). Thus, persons living with HIV are impaired on this semi-naturalistic task, despite the implicit opportunity to utilize effective compensatory mnemonic strategies. We have previously argued that limited metacognitive awareness (i.e., a deficit in “metaProM”) may at least partly explain the relationship between HIV-associated ProM impairment and problems in daily functioning (Woods et al., 2007, in press^a). Self-reported and objective ProM are fairly weakly related in HIV (i.e., Spearman’s rho = −0.15; Woods et al., 2007), meaning that individuals with HIV infection may misperceive their actual ProM abilities. In the context of medication adherence, a person with limited awareness of their ProM deficit (i.e., high expectations of their future ProM performance, despite prior evidence of impairment) may not employ otherwise effective compensatory strategies (e.g., a pillbox) and thereby be at risk for mismanaging their medication regimen. The likely complex relationships among discrepancy of expected and actual ProM, the use of compensatory strategies, and functional outcomes (e.g., medication nonadherence) is an important avenue for further research.

Consistent with prior studies in HIV (e.g., Hinkin et al., 2002; 2004), standard clinical measures of executive functions (i.e., planning and divided attention), attention, episodic retrospective memory, and verbal fluency were also predictive of medication management in this sample. More importantly, HIV-associated ProM impairment emerged as a significant predictor of medication management, even after considering the contributions of these established cognitive domains. In other words, ProM deficits explained a considerable amount of variance in medication management (i.e., 30%) beyond that which was captured by a traditional battery of clinical tests known to be sensitive to ARV nonadherence. This finding converges with a recent study showing that ProM demonstrates incremental ecological validity as a predictor of general IADL dependence in HIV (Woods et al., in press^a). Together, these results support the hypothesis that ProM captures a unique and largely untapped aspect of cognition that is highly germane to numerous aspects of successful daily living, perhaps most notably medication adherence (Park & Kidder, 1996).

Of course, neuropsychological impairment is not the sole determinant of medication nonadherence; indeed, although unique and robust, the overall magnitude of the direct relationship between ProM and medication management was relatively small. As in prior research (e.g., Barclay et al., 2007; Reynolds et al., 2004; Wilson et al., 2004), ethnicity, affective distress, psychosocial variables, and environmental factors were also predictive of medication management in this cohort. Nevertheless, HIV-associated ProM impairment remained a significant, independent contributor to poorer medication management vis-à-vis these well-established non-cognitive factors. It therefore stands to reason that assessment of ProM functioning may provide an added degree of predictive power to the standard comprehensive clinical evaluation of medication nonadherence risk. Studies are needed examine incremental validity of ProM in comparison to other known predictors of nonadherence in HIV, including self-efficacy, social support, medication side effects, socioecononic status, and various systemic factors (e.g., access to healthcare).

The primary limitation of this study was the exclusive reliance on a self-report questionnaire of medication management, which may be less sensitive than other approaches, such as pill counts, electronic monitoring devices, and analysis of blood levels (see Levine et al., 2006). In addition, the BERMA MMES is a general measure of medication management that is not specific to ARVs, which (along with the relative immunocompetence of the study sample) may explain its nonsignificant relationship to HIV disease variables. Nevertheless, the BERMA’s lack of specificity is not entirely disadvantageous because HIV-infected persons are often prescribed nonARV medications that are not typically considered in studies of ARV adherence. Nevertheless, because there is no universally accepted gold standard to measuring medication adherence in HIV, future studies may wish to use a multimodal longitudinal evaluation of ARV (and nonARV) adherence. Another limitation concerns the generalizability of these data, which were derived from a highly educated, mostly Caucasian male cohort with fairly well controlled HIV disease. Moreover, no participants had current substance-related disorders, which prior studies suggest are a significant risk factor for nonadherence (e.g., Wilson et al., 2004; cf. Waldrop-Valverde et al., 2004). Clarifying the role of ProM impairment in HIV-infected persons who are active substance users is therefore an important direction for future research, especially in the context of methamphetamine dependence, which is associated with increased rates of nonadherence (Reback et al., 2003) and memory impairment (e.g., Rippeth et al., 2004).

Regarding the potential clinical value of these findings, formal assessment of ProM abilities and complaints may be an invaluable component of evaluating the daily functioning capacity in HIV. Steady progress in the development of user-friendly clinical tests of ProM with robust psychometric properties, demographically-adjusted normative standards, and strong evidence of construct validity is critical in this regard. From a rehabilitation perspective, research is needed on the effectiveness of ProM-based interventions for improving medication adherence. The profile of HIV-associated ProM impairment may be most successfully confronted with remedial strategies that increase the quality, distinctiveness, and relatedness of external cues, which research shows enhances ProM, even in the setting of heightened cognitive load (McDaniel & Einstein, 1993; 2007). For example, a programmable electronic device that prominently notifies the patient when it is time to take a medication with a detailed text message that includes the medication, dosage, and particular conditions under which it should be taken (e.g., with food) might be maximally effective (e.g., Andrade et al., 2005; Leirer et al., 1991). Other empirically supported and potentially useful strategies for conditions in which such external cues are impractical might include explicit reminders from significant others or caregivers, minimizing internal and external distractions, increasing strategic cue monitoring, and implementing spaced-retrieval techniques (see McDaniel & Einstein, 2007).