Research report

Seasonal variation in suicidal behavior with prescriptionopioid medication

Jonathan M. Davis a,c,n,1, Veronica B. Searles b, S. Geoff Severtson a, Richard C. Dart a,c,Becki Bucher-Bartelson a

a Rocky Mountain Poison and Drug Center, Denver Health and Hospital Authority, 777 Bannock St., MC 0180, Denver, CO 80204, United Statesb University of Colorado Denver School of Medicine, Medical Scientist Training Program, United Statesc University of Colorado Denver School of Medicine, United States

a r t i c l e i n f o

Article history:Received 4 October 2013Accepted 11 January 2014Available online 31 January 2014

Keywords:Suicide attemptsSeasonalityPrescription opioidsAccessibility

a b s t r a c t

Background: Suicide attempts by self-poisoning utilizing prescription opioids account for more than halfof prescription drug abuse and misuse emergency calls received by poison centers. Additionally seasonalsuicidal behavior using other means is a commonly replicated finding. We hypothesized seasonalbehavior would exist in individuals using opioid medication as a suicide means, and that this patternwould change at different latitudes in the United States.Methods: We used a harmonic regression strategy to investigate sinusoidal seasonal variations of suicidalbehavior utilizing prescription opioids, and to contrast changes in seasonal behavior by latitude withinthe United States. Further, we investigated associations between suicide frequency utilizing opioidmedication and frequency of dispensed opioid prescriptions.Results: Seasonal patterns were identified; overall, all harmonic terms were significant (po0.05).Interaction terms of harmonic by latitude and harmonic by gender also were significant (po0.05). Afterstratification, females had significant harmonic terms at all latitudes. A changing peak time period withlatitude also was observed, such that the peak appeared later at more southern latitudes. Additionally,increased dispensed prescriptions rates per population were associated with increased suicidal behaviorutilizing prescription opioids.Limitations: This study has limitations due to its ecological nature and to missing data that may informour understanding of suicide risk factors, such as marital status and socio-economic status.Conclusion: Suicidal behavior with prescription opioids follows a seasonal pattern that changes withlatitude within the United States. Further, increased accessibility may contribute to increased suicidalattempt rates utilizing prescription opioids.

& 2014 Elsevier B.V. All rights reserved.

1. Background

1.1. Prescription opioids and suicide

Prescription opioid abuse has been increasing in recent yearsand has been a topic of major concern throughout the UnitedStates. Emergency department visits resulting from prescriptiondrug overdose have increased 111% from 2004 to 2008 (Cai et al.,2010), deaths increased three fold from 1999 to 2006 (Warneret al., 2009), and 8.6 billion dollars have been attributed to opioid

abuse through related medical expenses, criminal justice, andcosts due to lost work (Birnbaum et al., 2006).

An important yet often overlooked component of the cost ofprescription drug abuse is suicide by self-poisoning using pre-scription opioid medications. Suicide by any means is the 10thleading cause of death in the United States and accounted for36,909 deaths in 2010 (Murphy et al., 2012), and more than500,000 emergency department visits in 2008 (Ting et al., 2012).A major mechanism of suicide is intentional self-poisoning byopioids. One in five suicides has evidence of opiate involvement,including heroin and prescription pain killers (Karch et al., 2006).Among older women the use of hydrocodone for suicide increasedby 67% between 2004 and 2009 (Centers for Disease Control andPrevention (CDC), 2010).

Access to prescription opioid medication may lower suicideattempt threshold. Previous research has consistently shown thatas accessibility to the means for suicide increases, the risk for

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/jad

Journal of Affective Disorders

0165-0327/$ - see front matter & 2014 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.jad.2014.01.010

n Corresponding author at: Rocky Mountain Poison and Drug Center, DenverHealth and Hospital Authority, 777 Bannock St., MC 0180, Denver, CO 80204, UnitedStates.

E-mail addresses: jackdavis.epi@gmail.com,jonathan.davis@ucdenver.edu (J.M. Davis).

1 Site address: 990 Bannock St., Denver, CO 80204, USA.

Journal of Affective Disorders 158 (2014) 30–36

suicide increases; Marzuk et al. (1992) found that regions withincreased prescription medication accessibility had higher rates ofsuicide by overdose (Marzuk et al., 1992). Additionally, it has beennoted that opioid abusers report increased rates of suicide ideationand attempt compared to non-abusers (Kuramoto et al., 2012).This is particularly concerning given reports of increased numberof opioid prescriptions dispensed over recent years (Kuehn, 2007).

1.2. Seasonality of suicide

In addition to means accessibility, seasonality has been identi-fied as a suicide risk factor as increased rates of suicide arereported during specific times of year. Reports from northernEurope, the United States, Australia and Taiwan identifiedincreased suicide incidence during late spring and early summer(Chew and McCleary, 1995; Lee et al., 2006; Lester and Frank, 1988;Petridou et al., 2002). Length of daylight (Hiltunen et al., 2011),unseasonable temperature increases (Tsai and Cho, 2012), and airpollution (Szyszkowicz et al., 2010) have been associated withsuicide as well.

While seasonal suicidal behavior is a well-replicated finding,the cause of this behavior remains elusive. Two main hypotheseshave emerged over the course of study that focus on biological orenvironmental factors: (1) suicide frequency could be affected bybiological triggers such as melatonin regulation (Maes et al., 1996),dysfunctional circadian rhythm regulation (Petridou et al., 2002)and atopic disorders (Timonen et al., 2004), or (2) seasonal suicidevariation could be marking cultural changes such as seasonalgatherings or culturally significant dates (Araki and Murata,1987). One report found that the method used to commit suicidevaries seasonally (Lester and Frank, 1988), suggesting that season-ality may reflect seasonal changes in accessibility of means, apotential cultural mediator. Gender also may play a role, given thatfemales, in general, employ less lethal means but attempt moreoften than males (Hawton, 2000; Lester and Frank, 1988). Asfemales have higher rates of seasonal affective disorder(Kurlansik and Ibay, 2012), female attempts may be one of thedrivers of the noted seasonal differences. Further research exam-ining suicide attempt means and case seasonality is necessary toclarify cultural and biologic components. Such clarification willallow for better design of prevention strategies.

1.3. Latitude

Multiple studies have found that latitude is associated withsuicide rates and may modulate the relationship between season-ality and suicide. Davis and Lowell (2002) identified an associationbetween increasing latitude and greater variations in suicide ratesacross time (Davis and Lowell, 2002). They concluded that agreater variation in light–dark cycle exerts an environmental effecton suicide rates. This theory is supported by studies conducted atextreme latitudes that have identified seasonal variations in ratesversus studies conducted at moderate to low latitudes that failedto identify such an association (Flisher et al., 1997). A study ofsuicides in Greenland found seasonal variations on both sides ofthe Arctic Circle but noted that these variations were morepronounced further north (Björkstén et al., 2009). Rosen et al.,1990 also postulated that variations in light–dark cycles exerted aneffect on suicide rates, specifically via negative effects on circadianrhythms and, in turn, mental health (Rosen et al., 1990).

2. Hypothesis

The current study aimed to elucidate the relationships amongsuicide attempt rates, means accessibility, seasonality and latitude

with using prescription opiates as the means for attempts. Thehypotheses were: (1) that incidence of intentional self-poisoningwith opioid medications would increase with increased accessi-bility to prescription opioids; (2) would follow a seasonal patternsimilar to those previously reported; and (3) this pattern wouldchange depending on latitude within the United States. Stratifyingby latitude and examining accessibility could inform on theinfluence biologic and cultural factors have on suicide attemptswith prescription opioids.

3. Methods

3.1. Ethics statement

All participant data are de-identified and The Colorado MultipleInstitutional Review Board approved this research.

3.2. Poison center exposure data

Suspected suicide attempt case data came from the ResearchedAbuse, Diversion and Addiction-Related Surveillance (RADARSs)System Poison Center Program (Cicero et al., 2007) (www.radars.org). The RADARS System Poison Center Program is a nonprofitprescription drug abuse and misuse surveillance program housedby the Rocky Mountain Poison and Drug Center, a division of theDenver Health Hospital Authority. Participating poison centersreceived spontaneous calls from caregivers, patients, and healthcare providers regarding these exposures. During the study period,and within the catchment regions described further below,RADARS System poison centers served 36 states and approxi-mately 62% of the United States population. RADARS System datawere used to analyze patterns over time of cases involving suicideattempts related to suspected intentional exposure to prescriptionopioids from the 4th quarter 2006 through the 4th quarter of 2011.

Participating poison centers use a standardized electronichealth record to document individual cases involving exposuresto drugs of interest, including opioid drug classes. Case records areuploaded weekly to a central database maintained by the RADARSSystem. Upon receipt, the databases undergo multiple qualitycontrol processes, including verification of the drug(s) involvedand ensuring the accuracy and appropriateness of coded exposurereason based on American Association of Poison Control Centers(AAPCC) guidelines (AAPCC National Poison Data System (NPDS)Reference Manual., 2007). Intentional suspected suicide is used toclassify individuals who exposed themselves to the substance ofinterest with the intent to commit suicide. Medical outcomes arealso defined per AAPCC guidelines. A no effect medical outcome isdefined as an instance in which the individual developed nosymptoms from the exposure. A minor effect medical outcome isdefined as an instance when the individual has minimally bother-some symptoms, and a moderate effect medical outcome indicatesthe individual needed some form of treatment. The medicaloutcomes become progressively more serious, including majoreffects and death, with a major effect medical outcome recordedwhen lifesaving interventions are necessary; and the most seriousmedical outcome is death.

3.3. Case inclusion criteria

Computer case records were reviewed retrospectively for casesmentioning an intentional suspected suicide involving exposure toa RADARS System opioid of interest. All of the suicide attemptcases were included regardless of outcome. The opioid exposuresincluded oxycodone, fentanyl, hydrocodone, hydromorphone,morphine, oxymorphone, methadone, buprenorphine, tramadol

J.M. Davis et al. / Journal of Affective Disorders 158 (2014) 30–36 31

and tapentadol. Cases included in the analysis were from poisoncenters in the RADARS System with consistent coverage from the4th quarter of 2006 to the 4th quarter of 2011. Cases originatedfrom three-digit zip codes in northern states, middle states andsouthern states. California cases were split into middle and south-ern latitudes (Fig. 1).

3.4. Unique Recipients of Dispensed Drug (URDD) and populationcovered

An important component of this study is the construction ofpertinent denominators that adjust for changing populations,different population sizes and availability of medications. To dothis, the population covered by poison centers and the UniqueRecipients of Dispensed Drug (URDD) measure were utilized asvariables in the analysis.

The URDD measure is a quarterly count of individuals who haveopioid prescriptions and has been used as a marker of drugavailability (Dart, 2009). Conceptually, URDD is an ecologicalvariable that denotes the amount of prescription opioid medica-tion available in a specified time period and specified geography.Thus for the purposes of this study, URDD is an estimate of suicidemeans accessibility. URDD is linked ecologically to cases by drugclasses, case three digit zip codes, and case call dates. Although aquarterly measure, URDD has fluctuations over time.

The population covered in the study was geo-coded to UnitedStates Census three digit zip codes of poison center callers. Thesize of the three digit zip code population value was corrected forpopulation growth over time as reported in the Census at the statelevel. For example, the population of the United States grew 9.7%from 2000 to 2010 and specific states had different growth rates.The covered population was corrected in each three digit zip codelinearly per month over the time period using state specificgrowth rates.

The URDD and population covered variables are highly corre-lated (r¼0.97). Intuitively, this can be thought of as a phenomenonwhere a larger population will have higher URDD than that of asmaller population. To correct for this effect, we used a variablereflecting URDD relative to population size for use in the accessi-bility analysis (described in detail below). We defined accessibility

simply as URDD per three-digit zip code divided by the populationof the three-digit zip code.

3.5. Statistical methods

Two analyses were undertaken to evaluate the impact ofaccessibility (URDD/Population) on suicide frequency and season-ality of suicidal behavior involving prescription opioid medica-tions. First, we developed a linear model using suicide/attempt asthe dependent variable and accessibility as the primary explana-tory variable with sex as a covariate. The interaction of sex andaccessibility was tested as well to evaluate differences between thesexes. Autocorrelation in this model was tested via the DurbinWatson statistic and was substantial. Due to time correlations weincorporated a correlation structure that assumed a higher corre-lation of values at points closer in time.

Second, a population suicide attempt rate negative binomialmodel, including harmonic terms to mark seasonal patterns, wasconstructed to evaluate seasonal suicidal behavior. This modelincluded URDD as a covariate to clarify the potential seasonaleffects URDD may exhibit. As a quarterly measure we did notexpect URDD to definitively account for monthly seasonal changes.However, given the well documented drops in suicide frequencyover the winter period, we did expect to detect seasonal changesinfluenced by URDD over the winter months.

To model seasonal changes, a spectral analysis was performedassessing the sinusoidal characteristics of suicide attempts invol-ving prescription opioids. Seasonality was modeled through devel-oped cosine and sine functions (Cryer and Chan, 2008).Commonly:

mt ¼ β0þβ1 cos ð2πf tþΦÞþβ2 sin ð2πf tþΦÞþeij

where β is the amplitude f is the frequency, t is the period, and Φis the phase of the curve. Additional harmonic terms allow thedetection of more than one peak during the year. The model aboveis developed further through the addition of orthogonal Fourierterms that specify a more complex sinusoidal pattern:

mt ¼ β0þβ1 cos ð2πf tþΦÞþβ2 sin ð2πf tþΦÞþβ3 cos ð2πf =2tþΦÞþβ4 sin ð2πf =2tþΦÞþeij

Fig. 1. Highlights participating 3 digit zip codes from 4th quarter 2006 to 4th quarter 2011.

J.M. Davis et al. / Journal of Affective Disorders 158 (2014) 30–3632

Two harmonic regression coefficients were constructed to identifypatterns occurring over a 12 month and 6 month period asfollows:

β1 cos ð2πt=12Þ;β2 sin ð2πt=12Þand

β3 cos ð2πt=6Þ;β4 sin ð2πt=6ÞThe multi-spectral analysis was conducted with suicide attemptssummed by month as the dependent variable, harmonic terms asthe primary explanatory variables and additional covariates thatmay have relationship with suicide frequency in a negativebinomial population rate regression.

Month, month2, sex, and all interactions of harmonic bylatitude, harmonic by sex, and harmonic by latitude by sex wereincluded, as well as interactions of month and month2 by latitude.The polynomial month2 allowed the detection of a plateau, or aleveling off of the rate over time. The linearized corrected threedigit zip code population was used as the offset variable. Auto-correlation was tested using the Durbin Watson statistic in the fullregression model, which included harmonics and additional termsand was found to be unsubstantial. Stratified analyses were carriedout where interaction terms were significant. Customized hypoth-esis tests within the regression analysis were used to test differ-ences between females and males. Analysis was conducted usingthe GLIMMIX procedure in SAS Enterprise Guide v 4.3 (SASEnterprise Guide, SAS Institute, Cary, North Carolina, USA).

4. Results

Table 1 shows the medical outcomes recorded in case calls toRADARS System poison centers from the fourth quarter of 2006 to

the fourth quarter of 2011. A total of 74,655 cases included areported suicide attempt, accounting for 55.3% of all unintendeduse exposure cases. Lifesaving medical intervention was needed in5868 (7.9%) cases, as noted by a medical outcome recorded asmajor effect, and 515 (0.7%) cases resulted in death. The medianreported age was 36 years (25th–75th percentile range, 25–47),with 4.4% of cases missing exact age information and 45,823(61.4%) of cases were female.

Table 1 displays the range of medical outcomes recorded bypoison center specialists during the study period and within thecatchment region. The outcomes are predominantly indicative ofsuicide attempts.

Increased accessibility was strongly associated with increasedsuicide attempt frequency (po0.0001). This relationship wassignificantly different between the sexes (p¼0.001) (Fig. 2).

Table 1Medical outcomes recorded in case calls to poison centers.

Medical outcome Number N¼74,655 Percent

Missing 90 0.12No effect 11,628 15.58Minor effect 24,359 32.63Moderate effect 21,089 28.25Major effect 5868 7.86Death 515 0.69Not followed 11,106 16.29

Fig. 2. Displays linear relationships of suicide related calls and accessibility.Fig. 3. Predicted suicide attempts (dotted lines) and counts (solid lines) in femalesand males in northern, middle, and southern latitudes.

J.M. Davis et al. / Journal of Affective Disorders 158 (2014) 30–36 33

In the population suicide attempt rate seasonal model theoverall effect of the first and second harmonic terms was sig-nificant (all po0.05), as was the interaction of the first harmonicand latitude (p¼0.01), suggesting both, that there is an existingseasonal pattern and that this pattern varies by latitude. Springand fall peaks were observed in all latitude groups, although theamplitude of such peaks varied among latitudes (Fig. 3). Thegender by first sin harmonic interaction also was significant(p¼0.01), suggesting a different seasonal pattern between gen-ders. The timing of peaks was similar between genders but theamplitude of all harmonic terms was greater in females (po0.001)at all latitudes (Fig. 3). The three level interaction of harmo-nic� sex� latitude was not significant but did trend towardsignificance (p¼0.058), suggesting that seasonal patterns maydiffer by latitude and by sex. Interactions of latitude and time(month and month2) also were significant (po0.001), suggestingthat the population rate of case calls received related to suicideexposures over time differed among latitudes.

Due to the significant interactions, further analyses werestratified by gender and by latitude (Table 2). Significant harmoniceffects were observed in males in the middle latitudes only, whilefemales had significant harmonic effects in all latitudes. Interest-ingly peak suicide attempt time periods differed in females bylatitude. As latitude decreased, the first peak in females occurredlater in the year: April was the first peak in the northern latitude,May in the middle latitude and September in the southernlatitude.

Time findings, month and the polynomial month2 were sig-nificant in the stratified analysis and were different depending onlatitude. In the northern latitude an increase in suicide attemptrates over time throughout the study period were seen (Fig. 2),however this is not significant in females (p¼0.17) or in males(p¼0.08). In the middle and southern latitudes, a changing ratewas observed over time for both genders (po0.001) (Fig. 3) withan increase to a plateau after 2009.

After controlling for URDD results were similar. The overalleffect of adding URDD as a covariate to the seasonal model wasnegligible, suggesting that the seasonal patterns detected are notartifacts of prescription patterns.

Table 2 displays the stratified beta coefficients for harmonicterms in the sinusoidal regression model. The overall effects of all

harmonic terms prior to stratification were significant as well asinteraction terms suggesting that differences in the seasonalpattern existed between latitudes and genders. Bold font high-lights significant associations (po0.05).

5. Discussion

This study aimed to examine three hypotheses: (1) intentionalself-poisoning with prescription opioids increases with increasedopioid accessibility; (2) this will follow a seasonal pattern, in linewith reports from previous studies and (3), this seasonal patternwill change depending on latitude within the United States. Inbrief, the results support the hypotheses, demonstrating a positiveassociation between accessibility and suicide related opioid poisoncenter calls, seasonal variations in attempt rates, latitude effects onseasonal and time variations and gender effects on the relation-ship between accessibility and suicide rates, on seasonal suicidepatterns, and on latitude trends.

5.1. Accessibility

The relationship between means accessibility and suicide ratehas been described previously for firearms (Wintemute et al.,1999), prescription medications in general (Marzuk et al., 1992),and automobile emissions (Studdert et al., 2010). Literature exam-ining accessibility has repeatedly noted that the removal of meansis correlated with reduced suicide frequency (Sarchiapone et al.,2011). For example, constructing barriers at jumping sites andrestricting access to firearms have been associated with reducedrates of suicide using those methods (Sarchiapone et al., 2011). Thefindings presented here are consistent with this research showingthat suicide rates are associated with means accessibility.Increased accessibility in the form of increased prescriptions notedby Kuehn (2007_ may explain the rising suicide attempt rates seenover the five year study period. However, this trend remainedwhen we adjusted for URDD suggesting that the increasesobserved, while associated with accessibility, are potentially dueto additional unmeasured factors. This time period also spans theeconomic downturn in the US in 2008 and thus socioeconomicfactors may partially explain the rate increase during the study

Table 2Stratified beta values for seasonal regression coefficients in Females and Males by latitude.

Females Males

North latitudeHarmonic/Time Beta se p-Value Harmonic time Beta se p-Valueβ1 cos �0.019 0.014 0.173 β1 cos �0.001 0.018 0.631β2 sin �0.060 0.014 o0.001 β2 sin �0.034 0.017 0.052β3 cos 0.035 0.014 0.012 β3 cos �0.011 0.017 0.520β4 sin 0.021 0.014 0.127 β4 sin �0.020 0.017 0.251Month 0.003 0.002 0.173 Month 0.005 0.003 0.079Month2 0.000 0.000 0.618 Month2 �0.000 0.000 0.748

Middle latitudeβ1 cos 0.025 0.013 0.057 β1 cos 0.035 0.016 0.025β2 sin �0.041 0.013 0.002 β2 sin 0.01 0.016 0.740β3 cos 0.002 0.013 0.850 β3 cos 0.032 0.016 0.044β4 sin 0.045 0.013 o0.001 β4 sin 0.021 0.016 0.172Month 0.012 0.002 o0.001 Month 0.016 0.003 o0.001Month2 �0.0001 0.00003 0.006 Month2 �0.0002 0.00004 o0.001

Southern latitudeβ1 cos 0.033 0.016 0.035 β1 cos 0.021 0.018 0.245β2 sin �0.043 0.016 0.005 β2 sin �0.006 0.018 0.734β3 cos 0.021 0.016 0.176 β3 cos 0.028 0.018 0.111β4 sin 0.007 0.015 0.651 β4 sin 0.019 0.018 0.275Month 0.019 0.003 o0.001 Month 0.023 0.003 o0.001Month2 �0.0002 0.00004 o0.001 Month2 �0.0003 0.00004 o0.001

J.M. Davis et al. / Journal of Affective Disorders 158 (2014) 30–3634

period. Future research should investigate the magnitude of socio-economic contributions to attempt rates involving prescriptionopioids.

The congruence between the data from this study and previousfindings suggests that means accessibility is a major contributingrisk factor for suicide. The data thus support the theory thatprimary care physician training and restricting access may be themost effective public health based suicide prevention strategieswith regard to opioid medications (Mann et al., 2005). Whileopioid medications are a vital component of pain management,increased prescriber training and prescription rate reduction couldaddress pain management while concurrently reducing the fre-quency of suicide attempts that use opioids as a means.

5.2. Seasonality and latitude

The data show a notable seasonal pattern in suicide attempt rates.This supports previous research that has identified seasonality ofsuicide rates with evidence of increased frequency in late spring andfall. Hypotheses put forth to explain seasonal variations includebioclimatic factors such as melatonin regulation (Maes et al., 1996)or seasonal cultural gatherings or economic changes over time (Arakiand Murata, 1987). The data demonstrate that seasonal suicidalattempts using prescription opioids differ by latitude: with decreasinglatitude, the first peak in rate happens later in the year. The differentharmonic patterns observed at different latitudes presented in thisreport may at first appear to suggest a bioclimatic effect, such asseasonal changes in pollen count or temperature variation. However,the seasonal changes occurring at different latitudes happen atdifferent time points than the seasonal suicide peaks noted here.Spring typically arrives earlier in the year in the south, yet the seasonalsuicide peak in the south occurred later in the year than otherlatitudes. This imbalance complicates bio-climatic effects that maybe related to seasonal changes.

It should be noted that cultural aspects including race, ethnicityand economic conditions may vary by latitude, and those effectscould influence a seasonal pattern at the population level. Ascultural data are not recorded emergency poison center calls, wecannot evaluate the effect of these variables. Future researchshould attempt to elucidate these potential effects on seasonalpatterns. Furthermore, the negligible effect of URDD as a covariatesuggests the seasonal pattern observed is not an artifact ofprescription frequencies that may be changing in time.

Wemust also address the fact that multiple studies have reported alack of seasonal effects or decreasing seasonal effects over time. Noseasonal effect was identified in Brazil (Nejar et al., 2007), a lack of aspring and summer effect was found inWales (Page et al., 2007), and asimilar lack of seasonality in England and Wales exists (Yip et al.,2000). The lack of seasonal effect observed in Wales may be due to itsgeographic location. Temperature and weather patterns are morehomogenous in maritime regions than in the continental regions weexamined, and seasonal effects may vary with extremity of weathervariations. For example, Page et al., 2007, remark on a suicideassociation with unseasonably increased temperature but not anassociation with a seasonal pattern. Likewise correlations of suicidewith sunlight and temperature change suggest that the seasonalphenomena may be driven mostly by factors related to the net changein weather. Perhaps the changing peaks observed in this report areindicative of changes in continental climates that exist progressivelynorthward, where, for example, the seasonal changes in Texas aremore subdued than those in Minnesota.

5.3. Gender

We found different patterns of seasonal suicide betweengenders using self-poisoning as a method. This may be expected

as males tend to use more violent means of suicide and malesuicides comprise 78% of total suicides (Karch et al., 2006), andthis analysis identified a majority of female calls (61%) due to self-poisoning. However, this study is powered adequately to detectpotential seasonal differences in males, yet only does so in themiddle latitude. This argues that the seasonal effects identified infemales in this study are due to a factor unique to females beyonda preference for self-poisoning. The RADARS System data does notinclude information on prescriptions of opioid medication differ-ences between genders and thus we cannot determine if thisdifference is based on differences in prescription rates alone.Disproportionate numbers of prescriptions for females mayexplain a portion of increased rates in females, but this is a largestudy adequately powered to detect trends in males, and in maleswe see minimal seasonal effects. Additionally, seasonal affectivedisorder (SAD) is more prevalent in females (Saeed and Bruce,1998), particularly atypical SAD, or summer depression. AtypicalSAD is notable for symptoms not seen in traditional SAD, includingagitation, insomnia and suicidal thoughts (Wehr et al., 1991). Thisstudy is not able to assess atypical SAD from poison center casecalls, but future studies incorporating SAD presence may providefurther information on factors contributing to suicidal behaviors inindividuals who use opioids as an attempt method.

5.4. Strengths and limitations

This study has limitations due to its ecological nature and tomissing data that may inform our understanding of suicide riskfactors, such as marital status and socio-economic status. Addi-tionally, the inclusion of URDD in the seasonal analysis does notallow a direct comparison to monthly prescription rates as URDDis a quarterly measure. However, no change in seasonal effects wasdetected with the inclusion of URDD as a covariate, and reportsregarding seasonal opioid prescription practices are lacking. Addi-tional investigation including other risk factors will allow moredetailed identification of a specific subgroup and inform ourunderstanding of the seasonality of suicide related to prescriptionopioid medication exposures.

This study also has numerous strengths. First is the timelinessof the findings: this study analyzed cases over a five-year periodending in 2011, and thus our conclusions are more applicable tocurrent prevention efforts than those from studies covering a shorttime span or examining an older case set. A second strength is thelarge catchment area covered, encompassing approximately 62% ofthe total US population and thus providing results that are morebroadly applicable than those from a smaller geographic region.Additionally, this study takes a unique approach to analyzingsuicide attempt cases, focusing specifically on self-poisoning withprescription opioid medications. Incorporating our results into thecurrent body of knowledge will aid future prevention measures byproviding better information regarding the factors that maypredict attempts in opioid users and critical time periods forintervention and means restriction.

Role of funding sourceThis was unfunded research. The authors (JD, SS, RD, and BB) were or currently

are employees of the Denver Health and Hospital Authority – Rocky MountainPoison and Drug Center (RMPDC).

Conflict of interestDenver Health and Hospital Authority – Rocky Mountain Poison and Drug

Center (RMPDC), a public non-profit organization, operates the Researched Abuse,Diversion and Addiction-Related Surveillance (RADARS(r)) System which providesdata to the pharmaceutical industry, regulatory agencies, and researchers. TheRADARS(r) System is funded by subscriptions to periodic reports, and industry andgovernment grants. Financial supporters did not participate in the development of

J.M. Davis et al. / Journal of Affective Disorders 158 (2014) 30–36 35

the research question, analytical methods, or the interpretation of the results forthis report. RMPDC employees receive only their salary for participation in systemoperations and research activities; employees have no direct financial relationshipwith any of the subscribers or grant providers.

AcknowledgmentAt this time we have no acknowledgments to make.

References

AAPCC National Poison Data System (NPDS) Reference Manual, 2007. Washington,DC.

Araki, S., Murata, K., 1987. Suicide in Japan: socioeconomic effects on its secular andseasonal trends. Suicide Life Threatening Behav. 17, 64–71.

Birnbaum, H.G., White, A.G., Reynolds, J.L., Greenberg, P.E., Zhang, M., Vallow, S.,Schein, J.R., Katz, N.P., 2006. Estimated costs of prescription opioid analgesicabuse in the United States in 2001: a societal perspective. Clin. J. Pain 22,667–676.

Björkstén, K.S., Kripke, D.F., Bjerregaard, P., 2009. Accentuation of suicides but nothomicides with rising latitudes of Greenland in the sunny months. BMCPsychiatry 9, 20.

Cai, R., Crane, E., Poneleit, K., Paulozzi, L., 2010. Emergency department visitsinvolving nonmedical use of selected prescription drugs in the United States,2004–2008. J. Pain Palliat. Care Pharmacother. 24, 293–297.

Centers for Disease Control and Prevention (CDC), 2010. Emergency departmentvisits involving nonmedical use of selected prescription drugs—United States,2004–2008. MMWR. Morb. Mortal. Wkly. Rep. 59, 705–709.

Chew, K.S., McCleary, R., 1995. The spring peak in suicides: a cross-nationalanalysis. Soc. Sci. Med. 40, 223–230.

Cicero, T.J., Dart, R.C., Inciardi, J.A., Woody, G.E., Schnoll, S., Muñoz, A., 2007. Thedevelopment of a comprehensive risk-management program for prescriptionopioid analgesics: researched abuse, diversion and addiction-related surveil-lance (RADARS). Pain Med. 8, 157–170.

Cryer, J.D., Chan, K.-S., 2008. Time Series Analysis: With Applications in R, seconded. Springer Science þ Business Media, LLC, New York, NY, USA.

Dart, R.C., 2009. Monitoring risk: post marketing surveillance and signal detection.Drug Alcohol Depend. 105 (Suppl 1), S26–S32.

Davis, G.E., Lowell, W.E., 2002. Evidence that latitude is directly related to variationin suicide rates. Can. J. Psychiatry 47, 572–574.

Flisher, A.J., Parry, C.D., Bradshaw, D., Juritz, J.M., 1997. Seasonal variation of suicidein South Africa. Psychiatry Res. 66, 13–22.

Hawton, K., 2000. Sex and suicide gender differences in suicidal behaviour. BJP 177,484–485.

Hiltunen, L., Suominen, K., Lönnqvist, J., Partonen, T., 2011. Relationship betweendaylength and suicide in Finland. J. Circadian Rhythms 9, 10.

Karch, D.L., Dahlberg, L.L., Patel, N., Davis, T.W., Logan, J.E., Hill, H.A., Ortega, L.,2006. Centers for Disease Control and Prevention (CDC), 2009. Surveillance forviolent deaths-national violent death reporting system, 16 States. MMWRSurveillance Summ. 58, 1–44.

Kuehn, B.M., 2007. Opioid prescriptions soar: increase in legitimate use as well asabuse. JAMA 297, 249–251.

Kuramoto, S.J., Chilcoat, H.D., Ko, J., Martins, S.S., 2012. Suicidal ideation and suicideattempt across stages of nonmedical prescription opioid use and presence ofprescription opioid disorders among U.S. adults. J. Stud. Alcohol Drugs 73,178–184.

Kurlansik, S.L., Ibay, A.D., 2012. Seasonal affective disorder. Am. Fam. Physician 86,1037–1041.

Lee, H.-C., Lin, H.-C., Tsai, S.-Y., Li, C.-Y., Chen, C.-C., Huang, C.-C., 2006. Suicide ratesand the association with climate: a population-based study. J. Affect. Disord. 92,221–226.

Lester, D., Frank, M.L., 1988. Sex differences in the seasonal distribution of suicides.Br. J. Psychiatry 153, 115–117.

Maes, M., Scharpé, S., D'Hondt, P., Peeters, D., Wauters, A., Neels, H., Verkerk, R.,1996. Biochemical, metabolic and immune correlates of seasonal variation inviolent suicide: a chronoepidemiologic study. Eur. Psychiatry 11, 21–33.

Mann, J.J., Apter, A., Bertolote, J., Beautrais, A., Currier, D., Haas, A., Hegerl, U.,Lonnqvist, J., Malone, K., Marusic, A., Mehlum, L., Patton, G., Phillips, M., Rutz,W., Rihmer, Z., Schmidtke, A., Shaffer, D., Silverman, M., Takahashi, Y., Varnik, A.,Wasserman, D., Yip, P., Hendin, H., 2005. Suicide prevention strategies: asystematic review. JAMA 294, 2064–2074.

Marzuk, P.M., Leon, A.C., Tardiff, K., Morgan, E.B., Stajic, M., Mann, J.J., 1992. Theeffect of access to lethal methods of injury on suicide rates. Arch. Gen.Psychiatry 49, 451–458.

Murphy, S., Xu, J., Kochanek, K., 2012. Deaths: Preliminary Data for 2010.Nejar, K.A., Benseñor, I.M., Lotufo, P.A., 2007. Sunshine and suicide at the tropic of

Capricorn, São Paulo, Brazil, 1996–2004. Rev. Saude Publica 41, 1062–1064.Page, L.A., Hajat, S., Kovats, R.S., 2007. Relationship between daily suicide counts

and temperature in England and Wales. Br. J. Psychiatry 191, 106–112.Petridou, E., Papadopoulos, F.C., Frangakis, C.E., Skalkidou, A., Trichopoulos, D.,

2002. A role of sunshine in the triggering of suicide. Epidemiology 13, 106–109.Rosen, L.N., Targum, S.D., Terman, M., Bryant, M.J., Hoffman, H., Kasper, S.F.,

Hamovit, J.R., Docherty, J.P., Welch, B., Rosenthal, N.E., 1990. Prevalence ofseasonal affective disorder at four latitudes. Psychiatry Res. 31, 131–144.

Saeed, S.A., Bruce, T.J., 1998. Seasonal affective disorders. Am. Fam. Physician 57(1340–1346), 1351–1352.

Sarchiapone, M., Mandelli, L., Iosue, M., Andrisano, C., Roy, A., 2011. Controllingaccess to suicide means. Int. J. Environ. Res. Public Health 8, 4550–4562.

Studdert, D.M., Gurrin, L.C., Jatkar, U., Pirkis, J., 2010. Relationship between vehicleemissions laws and incidence of suicide by motor vehicle exhaust gas inAustralia, 2001–06: an ecological analysis. PLoS Med. 7, e1000210.

Szyszkowicz, M., Willey, J.B., Grafstein, E., Rowe, B.H., Colman, I., 2010. Air pollutionand emergency department visits for suicide attempts in Vancouver, Canada.Environ. Health Insights 4, 79–86.

Timonen, M., Viilo, K., Hakko, H., Särkioja, T., Meyer-Rochow, V.B., Väisänen, E.,Räsänen, P., 2004. Is seasonality of suicides stronger in victims with hospital-treated atopic disorders? Psychiatry Res. 126, 167–175.

Ting, S.A., Sullivan, A.F., Boudreaux, E.D., Miller, I., Camargo , C.A., 2012. Trends in USemergency department visits for attempted suicide and self-inflicted injury,1993–2008. Gen. Hosp. Psychiatry 34, 557–565.

Tsai, J.-F., Cho, W., 2012. Temperature change dominates the suicidal seasonality inTaiwan: a time-series analysis. J. Affect. Disord. 136, 412–418.

Warner, M., Chen, L.H., Makuc, D.M., 2009. Increase in fatal poisonings involvingopioid analgesics in the United States, 1999–2006. NCHS Data Brief, 1–8.

Wehr, T.A., Giesen, H.A., Schulz, P.M., Anderson, J.L., Joseph-Vanderpool, J.R., Kelly,K., Kasper, S., Rosenthal, N.E., 1991. Contrasts between symptoms of summerdepression and winter depression. J. Affect. Disord. 23, 173–183.

Wintemute, G.J., Parham, C.A., Beaumont, J.J., Wright, M., Drake, C., 1999. Mortalityamong recent purchasers of handguns. N. Engl. J. Med. 341, 1583–1589.

Yip, P.S., Chao, A., Chiu, C.W., 2000. Seasonal variation in suicides: diminished orvanished. Experience from England and Wales, 1982–1996. Br. J. Psychiatry 177,366–369.

J.M. Davis et al. / Journal of Affective Disorders 158 (2014) 30–3636