This is an open-access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included.
Krokodil is an informal term for a cheap injectable illicit drug domestically prepared from codeine-containing medication (CCM). The method of krokodil preparation may produce desomorphine as well as toxic reactants that cause extensive tissue necrosis. The first confirmed report of krokodil use in Russia took place in 2004. In 2012, reports of krokodil-related injection injuries began to appear beyond Russia in Western Europe and the United States.
This exploratory study had two main objectives: (1) to determine if Internet search patterns could detect regularities in behavioral responses to Russian CCM policy at the population level, and (2) to determine if complementary data sources could explain the regularities we observed.
First, we obtained krokodil-related search pattern data for each Russia subregion (oblast) between 2011 and 2012. Second, we analyzed several complementary data sources included krokodil-related court cases, and related search terms on both Google and Yandex to evaluate the characteristics of terms accompanying krokodil-related search queries.
In the 6 months preceding CCM sales restrictions, 21 of Russia's 83 oblasts had search rates higher than the national average (mean) of 16.67 searches per 100,000 population for terms associated with krokodil. In the 6 months following restrictions, mean national searches dropped to 9.65 per 100,000. Further, the number of oblasts recording a higher than average search rate dropped from 30 to 16. Second, we found krokodil-related court appearances were moderately positively correlated (Spearman correlation=.506,
Illicit drug use data are generally regarded as difficult to obtain through traditional survey methods. Our analysis suggests it is plausible that Yandex search behavior served as a proxy for patterns of krokodil production and use during the date range we investigated. More generally, this study demonstrates the application of novel methods recently used by policy makers to both monitor illicit drug use and influence drug policy decision making.
Krokodil, otherwise known as desomorphine, is a cheap injectable drug easily synthesized in household kitchens from codeine-containing medication (CCM). The first confirmed report of krokodil use in Russia occurred in 2004. In 2012, reports of horrific krokodil-related injection injuries began to appear beyond Russia in Western Europe [
Current scientific literature on krokodil is limited. We reviewed international literature available through PubMed and Google Scholar. In addition, we searched the four most popular Russian online news sources [
Russian news sources reviewed via Yandex News (Jan 1, 2009 to Dec 31, 2012).
Source | Website | Count (N=929) | Orientation |
RIA Novosti | rian.ru | 103 | State-owned |
Vesti.ru (Website of Russia 24 TV) | vesti.ru | 38 | State-owned |
Komsomlskaya Pravda | kp.ru | 748 | Private, tabloid |
Russia business consulting | rbk.ru | 40 | Private, business focus |
Current literature describing the origins of krokodil in Russia is vague. Time magazine reported the first appearance of krokodil in the Siberian and the Far East Federal Regions of Russia in the early 2000s [
Estimates of the scale of krokodil use diverged markedly. In 2011, a senior Russian addiction medicine specialist reported 5000 krokodil users were receiving treatment nationally, out of a total estimated national population of 20,000-30,000 users [
There were limited data describing the spatial distribution of krokodil use before the federal restrictions on sales of CCM in June 2012. Krokodil use had been widely reported across Russia and bordering Ukrainian regions [
A series of articles reviewing the current scientific understanding of krokodil appeared in the International Journal of Drug Policy in 2013. One of these articles by Grund et al consolidated current scientific information, including interview data from Russian key informants [
Desomorphine was originally developed as a morphine substitute. It was first synthesized in the United States in 1932, with the aim of producing a low-cost substitute with minimal side effects [
Frequent injecting is generally regarded as a risk factor for human immunodeficiency virus (HIV) and other injecting-related harms [
The easing of restrictions on access to CCM may have increased the production and use of krokodil in Russia. During the Soviet period up to 1991, CCM was available only through pharmacies with a medical prescription [
At a national drug control conference in 2011, Russian President Medvedev announced restrictions on the sale of CCM without medical prescriptions [
Public opinion survey into consequences of proposed federal CCM restrictions May 2011 [
|
How will this affect the battle against drugs in Russia? | How will this affect the needs of ordinary patients? |
Generally positive | 32% | 11% |
No effect | 49% | 21% |
Generally negative | 5% | 56% |
Difficult to answer | 14% | 12% |
From 2011 to June 2012, several Russian oblast governments implemented interim local restrictions on CCM sales [
Between 2010 and 2012, Russian policy makers emphasized the negative influence of the Internet in disseminating krokodil-related information. In April 2011, the FSKN presented the results of its research into Internet search patterns for krokodil-associated terms [
President Medvedev’s demonstration stimulated increased public interest in krokodil. The FSKN had reported steadily increasing Internet searches in the 12 months before President Medvedev’s speech. However, extensive media coverage and the highest recorded volume of searches for “desomorphine” emerged in the week following the President’s speech (see
Political and public concern over illicit drug use preceded President Medvedev’s speech. Public opinion polls since 2005 consistently rated illicit drug use as one of the most serious social problems in Russia [
Desomorphine searches Google Trends Russia - 2009-2013.
Exposure to stories about illicit drugs in traditional and online media has been found to increase public curiosity and the use of illicit drugs [
We conducted this exploratory infodemiology study in order to better understand if the relative scale and spatial distribution of search behavior was consistent with an interest in the production and use of krokodil in Russia before and after the imposition of federal restrictions on CCM sales in 2012. In conducting this study, we examined “the science of distribution and determinants of information...(on) the Internet (and) a population, with the ultimate aim to inform public health and public policy” [
This study had two main objectives: (1) to determine if Internet search patterns could detect regularities in behavioral responses to Russian CCM policy at population level, and (2) if complementary data sources could explain the regularities we observed.
Each Internet search is a behavioral measure of an issue’s importance to an individual [
Most Internet search pattern studies have used Google Trends as the data source. Google Trends has been deployed in studies of influenza [
Illicit drug use data are generally regarded as difficult to obtain. Drug use estimates are imperfect even in high-income countries with adequate resources [
In addition, we identified two references to national drug agencies using novel methods for estimating illicit drug use prevalence. The UK government made use of Google Trends when considering restrictions on the novel substance mephedrone in 2010 [
To answer this question, we examined search patterns using Yandex and Google Internet search engines. First, we determined the most appropriate search term to represent the informal term “krokodil” in Internet searches. We initially selected two search terms that we believed reflected the majority of searches for the concept central to this study [
Google Trends related terms for desomorphine, Russian Federation.
Search terms | Russian | Value | |
|
|||
|
Desomorphine how to prepare | дезоморфин как приготовить | 100 |
|
Prepare desomorphine | приготовить дезоморфин | 100 |
|
Krokodil desomorphine | крокодил дезоморфин | 80 |
|
Krokodil | крокодил | 80 |
|
Drug desomorphine | наркотик дезоморфин | 60 |
|
Desomorphine recipe | дезоморфин рецепт | 50 |
|
Krokodil drug | крокодил наркотик | 30 |
|
|||
|
Gena the crocodile (Children’s animation) | крокодил гена | 100 |
|
Crocodile game (Children’s game) | игра крокодил | 50 |
|
Krokodil drug | крокодил наркотик | 50 |
|
Crocodile/krokodil online | крокодил онлайн | 45 |
|
Crocodile Dundee (Australian film) | крокодил данди | 35 |
|
Dundee (Australian film) | данди | 35 |
|
Cheburashka (Children’s animation) | чебурашка | 30 |
Publicly available Google Trends data for Russia has several limitations. First, Google did not provide complete results, returning only oblasts with the highest search volume. Google data for the term desomorphine was available for only 8 of Russia’s 83 oblasts and 3 cities during the date range 2011-2013. Second, Google did not provide raw search data. This made direct comparisons between oblasts using Google data impossible. We thus used WordStat as the primary data source. Yandex made publicly available a complete raw search dataset for all Russian regions and oblasts for 6 months before and after the implementation of federal CCM restrictions in June 2012. We used Google Trends as a secondary source of aggregated search results for validation purposes.
Second, we obtained desomorphine search data for each Russia oblast from September 1, 2011, to August 31, 2013. Yandex provides 2 years of publicly available monthly search pattern data at any time. Additionally, we had 6 months previously downloaded monthly search pattern data for the term desomorphine for each Russian oblast, from February to August 2011.
Third, we converted raw search figures for the term desomorphine to population prevalences. This allowed direct comparison across regions and oblasts. We used 2010 federal Russian census data [
Fourth, we analyzed search patterns before and after federal restrictions on CCM sales in June 2012. We obtained the mean search volume for 6 months before the restrictions, as well as 6 and 12 months after (ie, to August 31, 2013). We excluded June 2012 data, as we anticipated atypical search patterns in the immediate post-restriction period. Overall, we segmented the available data to examine the effects of a federal policy change on the relative scale and geographic patterns of krokodil search across the Russia.
Fifth, we obtained all available Google data for the term desomorphine from September 2011 to August 2012. Google search data for the term desomorphine was available for 8 of 83 oblasts only (see
All available Google Trends results in Russia from September 2011 to September 2013.
Region | Search volume | |
|
||
|
Chelyabinsk | 100 |
|
Novosibirsk | 88 |
|
Sverdlovsk | 86 |
|
Samara | 85 |
|
Rostov | 83 |
|
Saint Petersburg city | 71 |
|
Moscow city | 68 |
|
Krasnodar | 67 |
|
||
|
Yekaterinburg | 100 |
|
Nizhny Novgorod | 98 |
|
Chelyabinsk | 89 |
|
Samara | 87 |
|
Novosibirsk | 87 |
|
Rostov-on-Don | 85 |
|
Saint Petersburg | 81 |
|
Moscow | 77 |
|
Kazan | 68 |
|
Krasnodar | 61 |
To answer this question, we initially reviewed the approaches used to validate search pattern data and drug population data. Search pattern studies have generally validated against an offline measure. For example, the initial search pattern studies established correlations between search patterns and epidemiological surveillance data for influenza [
First, we obtained first court appearance data available for krokodil-related criminal charges for the 77 of 83 Russian oblast data available from the Rospravosudie website. The site is a publicly available, non-government Russian criminal justice research project displaying criminal court case data across all Russian oblasts [
To analyze krokodil-related court data, we first obtained arrest rates for krokodil for 2010-2012 as a single figure for each Russian oblast. We then converted the arrest rates for each oblast to a per 100,000 population measure. This allowed us to investigate the relationship between court appearances and krokodil searches. We used the mean searches for “desomorphine” per 100,000 population from November 2011 to May 2012 to represent pre-CCM restriction searches. We then conducted Spearman correlation between arrest rates and searches per 100,000 population for “desomorphine” for the 77 regions for which court data were available.
Second, we used Google Trends visual data to provide indicative national search results for popular CCMs and “desomorphine” from January 2009 to January 2013. We identified several popular CCM available in Russia prior to the June 2012 ban [
Third, we used Google Trends related searches to analyze several popular CCM available in Russia prior to the June 2012 restrictions. Through analyzing these related searches, we sought to obtain additional information on the characteristics of public interest in CCM and the term desomorphine before and after federal restrictions. Historical Yandex data were not available for this complete date range.
Fourth, we used Yandex keyword feature nationally to confirm that searches for the term desomorphine were associated with illicit drug use. Yandex provides a keyword function that lists word combinations associated with a specified search term. Keywords are analogous to the Google related terms (Top Searches) feature [
In the 6 months before the CCM restrictions in June 2012, 21 of Russia’s 89 oblasts had Internet search rates higher than the national average (mean) of 16.67 per 100,000 (see
Yandex search patterns for “desomorphine” in selected Russian subregional cities.
Cities | Pre-ban, 6 months, Dec 2011-June 2012 | Post-ban, 6 months, July-Dec 2012 | Post-ban, 6 months, Feb-Sept 2013 | % change post-ban, 6 months | % change, Feb-Sept 2013 | |
|
||||||
|
Vologda city | 80.916 | 28.389 | 33.857 | 64.915 | 58.157 |
|
Cherepovets | 50.324 | 28.337 | 25.829 | 43.690 | 48.674 |
|
||||||
|
Yekaterinburg | 16.200 | 10.372 | 10.261 | 35.976 | 36.662 |
|
Kamensk-uralskiy | 20.226 | 13.262 | 6.583 | 34.434 | 67.453 |
|
Pervouralsk | 13.116 | 6.290 | 6.157 | 52.041 | 53.061 |
|
||||||
|
Rostov-na-donu | 56.460 | 43.409 | 42.353 | 23.117 | 24.986 |
|
Kamensk-Shakhtinsky | 17.850 | 13.002 | 5.950 | 27.160 | 66.667 |
|
Shakhty | 2.778 | 0.764 | 1.875 | 72.500 | 32.500 |
|
Volgodonsk | 26.633 | 11.219 | 8.780 | 57.875 | 67.033 |
|
Taganrog | 13.324 | 6.080 | 8.861 | 54.369 | 33.495 |
|
Novocherkassk | 6.420 | 4.247 | 3.852 | 33.846 | 40.000 |
|
||||||
|
Samara city | 25.329 | 16.442 | 24.227 | 35.085 | 4.350 |
|
Togliatti | 30.131 | 18.736 | 15.147 | 37.817 | 49.731 |
|
||||||
|
Sochi | 16.068 | 9.320 | 8.495 | 41.994 | 47.130 |
|
Novorossiysk | 11.986 | 11.159 | 3.582 | 6.897 | 70.115 |
|
Krasnodar city | 32.774 | 20.917 | 14.586 | 36.177 | 55.495 |
To answer this question, we used several complementary sources of krokodil-related data. We found a Spearman correlation of .506 (
Correlation between searches for the term desomorphine and court appearances.
No. of subregions | Data source | Date range | Spearman correlation |
83 (77 correlated) | “desomorphine” searches, Yandex WordStat | Dec 2011-May 12 | — |
77 | desomorphine court appearances | 2010-2012 | .506 ( |
Second, we examined national Google Trends results for four CCMs and “desomorphine”. Overall, search volumes for both CCM decreased in the 6 months before the June 2012 federal restrictions, as did searches for the term “desomorphine”. Public interest in CCM and the term desomorphine was roughly similar in the 6 months before the implementation of restrictions. The exception was an increase in search for the CCM pentalgin immediately before the June 2012 restrictions (see
Third, we examined Google Trends related terms for CCMs and desomorphine. We found related terms for CCMs consistent with therapeutic and analgesic uses (see
Google related search terms for “desomorphine” from 2009-2013.
Date range | Pentalgin (пенталгин) | Value | Codelac (коделак) | Value | Desomorphine (дезоморфин) | Value |
2009-2013 | Pentalgin N | 100 | Codelac broncho | 100 | Desomorphine how to prepare | 100 |
Pentalgin instructions | 65 | Codelac phyto | 75 | Desomorphine krokodil | 75 | |
Pentalgin composition | 50 | Codelac instructions | 75 | Krokodil | 70 | |
Pentalgin price | 45 | Codelac price | 65 |
|
|
|
Nurofen | 40 | Codelac syrup | 60 |
|
|
|
Dec 2011-May 2012 | Pentalgin N | 100 | Codelac phyto | 100 | How to prepare desomorphine | 100 |
Pentalgin instructions | 85 | Codelac instructions | 100 | Krokodil | 55 | |
July 2012-Aug 2013 | Insufficient search volume | Nil | Insufficient search volume | Nil | Insufficient search volume | Nil |
Fourth, we used the Yandex keyword feature to analyze the word combinations used with the search term desomorphine. We found combinations associated with krokodil preparation and use accounted for 46.613% of searches, images, and general information for 24.175%, and ambiguous terms for 29.212% (see
The preparation and use category included all terms associated with drug preparation and use. Images and entertainment included visual material and terms unlikely to be associated with drug use and preparation (eg, “YouTube desomorphine”, “junkies desomorphine”). In summary, we found the combination of search patterns with complementary methods useful for identifying behaviors consistent with an interest in the production and use of krokodil.
Main themes identified in WordStat keyword combined word searches for “desomorphine” (excluding non-combined word searches for the single term “desomorphine”.
Code |
|
n (N=6338) | Percentage |
1 | Preparation & Use | 2952 | 46.613 |
2 | Images & information | 1531 | 24.175 |
3 | Ambiguous | 1850 | 29.212 |
We found federal CCM restrictions in June 2012 coincided with changes in the relative scale and spatial patterns of Internet search behaviors consistent with an interest in the production and use of krokodil. These changes in Internet search appeared consistent with behaviors that may be anticipated in the production and use of krokodil in response to changed access to CCM.
We observed marked reductions in searches for the term desomorphine following CCM sales restrictions in June 2012. By comparison with the 6 months preceding federal restrictions, searches dropped by 42.095% nationally (see
Third, we found the Google data available were inadequate for statistical analysis. Insufficient Google Trends data were available to conduct statistical analysis to identify oblasts where krokodil use may be prevalent. Google Trends data were available for only 8 of 83 regions (see
We identified several complementary data sources that provided a plausible explanation for the observed regularities in Internet search data. First, we found a moderately strong positive correlation (Spearman correlation=.506) between the geographic distribution of court appearances for krokodil-related charges, and Internet searches for the term desomorphine. This result should be treated with some caution. Court appearance data were available for 78 of 83 statistical regions. This may have affected the strength of correlations. More significantly, international researchers generally regard Russian policing as predatory and beyond the rule of law [
Second, the available Google Trends data suggested public interest in CCM and the term desomorphine was roughly similar in the 6 months prior to federal restrictions. However, the searches for CCM and desomorphine-related terms were not identical. The interest in CCM and in desomorphine manifested as different national level search patterns over the date range. While we had insufficient Google data to conduct correlations, this difference is evident on visual inspection (see
Yandex keyword analysis revealed a consistent pattern of behavior that was consistent with an interest in the production and use of krokodil. Yandex keyword data also revealed a strong popular interest in visual images of desomorphine use (see
In summary, we used complementary data sources in order to investigate behaviors consistent with an interest in the production and use of krokodil. Our analysis suggests that these combined complementary sources, including online news sources, provided a useful addition to the conventional approaches used to analyze krokodil use in Russia. Further, our analysis also suggests it is plausible that Yandex search behavior served as a proxy for krokodil production and use in the date range 2011-2012.
Google Trends results for CCM and desomorphine search 2009 - 2013.
Our research suggests that further research into the use of search patterns for investigating illicit drug use prevalence is warranted. First, search patterns offer researchers and non-government groups an additional source of indirect data with which to track the prevalence of traditional and emerging synthetic drugs at low cost and in near real time. We identified two references to national drug agencies in United Kingdom and Russia using Internet search methods to research patterns of illicit drug use [
Second, the krokodil case represents an example of a broader class of illicit drug policy events. International and Russian researchers have partially attributed increased use of krokodil to decreased heroin supply after 2009. Similarly, in 2012, government policy blocked easy access to CCM. In each case, existing networks of PWID were disrupted, and patterns of illicit drug use rapidly changed [
Third, media censorship is increasing in contemporary Russia. However, our analysis of online information relied on measures of unobserved population level demand for online information only. By contrast, censorship may be expected to influence the supply of illicit drug-related information. Russian government actions restricting the supply of illicit drug information are well documented in international literature (eg, [
Finally, search methods do not estimate actual drug user population size. However, our research suggests search methods can complement existing drug-using population estimation methods. For example, the Yandex keywords feature potentially provides a novel data source with which to track monthly shifts in keywords for illicit drug–related terms. Keywords measures provide a low-cost method for identifying spatial shifts in the relative scale of public interest in terms that are consistent with an interest in the production and use of novel and emerging illicit drugs in an increasingly complex environment where the opportunities for conventional field work and surveys in Russia for international researchers are decreasing.
Illicit drug use data are generally regarded as difficult to obtain through traditional survey methods. We used complementary methods to explain observed regularities in patterns of Internet search behavior before and after the imposition of Russian federal restrictions on CCM sales in 2012. Our analysis suggests it is plausible that Yandex search behavior served as a proxy for patterns of krokodil production and use during the date range we investigated. More generally, this study demonstrates the application of novel methods recently used by policy makers to both monitor illicit drug use and influence drug policy decision making.
Yandex Wordstat Keywords: combinations of words searched for with "desomorphine", November 2013.
Per-capita mean searches for "desomorphine" in all Russian federal regions & subregions (**= region or subregion above national mean).
Analysis of regional Internet search patterns.
Illicit drug search popularity of related terms in Google Trends 2009-2014.
Google searches for popular illicit drugs in the Russian Federation 2009-2014.
codeine-containing medication
human immunodeficiency virus
people who inject drugs
Russian Federal Drug Control Service
The work of Peter Meylakhs was funded by the Basic Research Program of the National Research University Higher School of Economics, Russia. The authors would also like to thank Ms Svetlana Chernova in Ukraine for her assistance with data collection and coding, Associate Professor James Gillespie at the Menzies Centre for Health Policy, University of Sydney for his advice and support, and Ms Anya Sarang from the Andrey Rylkov Foundation in Moscow for facilitating contacts within Russia.
None declared.