Deterministic Matching Research Articles

Abstract W P312: Pre-Hospital and Hospital Data Linkage in an Organized Stroke System of Care: A Pilot Study

Objective: Quality improvement within organized stroke systems of care, where policies such as diversion to primary stroke centers (PSC) integrate pre-hospital strategies with hospital performance and patient outcomes, are dependent on accurate linkage of emergency medical services (EMS) and hospital patient data. Currently, such linkage is difficult due to patient privacy concerns, heterogeneous and discrete databases, and variability in data collection. The objective of this pilot project was to create and validate a linkage algorithm for EMS and hospital data from 2 primary stroke centers (PSC) in Chicago. Methods: A retrospective analysis of the Get With The Guidelines (GWTG) Stroke registry identified confirmed stroke patients treated at 2 regional PSCs during the first 2 months of 2013. Data from the EMS SafetyPAD database on all patients arriving to those 2 PSCs during the same period were also retrieved. These datasets were provided in CSV format without patient identifiers and inserted it into the Oracle Database for schema matching. We then used a deterministic matching process using 4 non-identifying variables (destination hospital, arrival date/time, age, and gender) in both pre-hospital (EMS) and hospital (GWTG) datasets to create an algorithm for probabilistic matching. A time difference of 30 minutes was used for arrival time to allow for the human error in the time documentation. The algorithm was first evaluated at PSC 1 then validated against data from PSC 2. Results: At PSC 1, there were 13 EMS arriving stroke patients during the study period. Of these, 11 were matched (84.6%) to EMS records using the deterministic matching algorithm. At PSC 2, there were 7 EMS arriving stroke patients during the same period, of which 5 (71.4%) were matched to EMS records. The overall matching rate was 80%. In the 4 non-matches, inaccurate date/time entry occurred. Conclusions: Implementing a pre-hospital to hospital data linkage system is feasible using 4 commonly included non-identifying data elements. Quality improvement initiatives such as decreasing scene-to-hospital time and assessing pre-hospital stroke assessment tools would be greatly facilitated by having these databases linked.

Trends in Reported Syphilis and Gonorrhea among HIV-Infected People in Arizona: Implications for Prevention and Control

HIV and sexually transmitted disease (STD) surveillance patterns in Arizona suggested the need for integrated data analyses to identify trends. We compiled all HIV/AIDS cases diagnosed from 1998 to 2008 that were reported in Arizona and syphilis or gonorrhea cases diagnosed from 1998 to 2008 in Arizona. We used deterministic matching to identify individuals who were diagnosed with HIV and one or more STDs, and calculated time intervals between diagnoses. Of 23,940 people with HIV/AIDS reported from 1998 to 2008, 1,899 (2.6%) had at least one syphilis or gonorrhea diagnosis from 1998 to 2008. Approximately 85% of these cases reported male-to-male sexual contact. Among males with syphilis, HIV coinfection increased from 0.5% in 1998 to 29.1% in 2008. Among males with gonorrhea, HIV coinfection increased from 2.0% in 1998 to 3.1% in 2008. Among HIV cases diagnosed from 2004 to 2008 and reported with at least one syphilis or gonorrhea diagnosis, the majority of syphilis cases (76.1%) were diagnosed at or after HIV diagnosis, whereas a majority of gonorrhea cases (54.9%) were diagnosed prior to HIV diagnosis. Use of the deterministic matching method identified increases in STD infections among HIV-infected people. The routine performance of this cross-matching method may be a useful tool in identifying these high-risk individuals so that targeted partner services and appropriate care referrals may be used in a timely fashion.

Time since last negative HIV test among men who have sex with men and people who use injection drugs in British Columbia, 2006-2011.

Canadian surveys of men who have sex with men (MSM) and people using injection drugs (IDU) demonstrate that most have tested for HIV at least once, but that half or fewer have done so in the previous year. To better inform targeted HIV testing guidelines for these populations, we derived estimates of inter-test interval (ITI) for persons newly diagnosed with HIV in British Columbia (BC) between 2006 and 2011, and assessed variables associated with longer ITI among MSM and IDU. Provincial HIV case report and testing data were linked by deterministic and probabilistic matching (based on unique personal health number, name, and date of birth). ITI was defined as time from last recorded negative to first positive HIV result; those with ITI ≤30 days were excluded. Of 2,004 eligible individuals, 1,116 (55.7%) had a recorded negative HIV test result in the previous ten years. Overall median ITI was 20 months with a skewed distribution (inter-quartile range 8-46); median ITI was 15 months for MSM and 21 months for IDU with 41.2% and 33.1% testing in the past year, respectively. Longer ITI was associated with older age for both groups, and among MSM with residence outside Vancouver and not known to have an HIV-positive partner. These findings highlight potential missed opportunities for earlier detection of HIV and prevention of secondary transmission among newly diagnosed MSM and IDU, and provide evidence to inform recommendations for HIV test frequency and testing strategies for these populations in BC.

Abstract WP233: Prehospital Stroke Screen Accuracy in a State with an Emergency Medical Services Protocol for Routing Patients to Acute Stroke Centers

Background: A growing number of states are enacting legislation or regulations that direct emergency medical service (EMS) routing to state designated or Joint Commission Certified Stroke Centers, based on prehospital stroke screens performed by EMS providers. Objective: Assess the sensitivity and specificity of the Cincinnati Prehospital Stroke Screen (CPSS) and the Los Angeles Prehospital Stroke Screen (LAPSS) when performed by EMS providers in the context of a state-wide EMS stroke patient transport protocol. Methods: We used a validated method of deterministic matching to link a statewide Prehospital Medical Information System database, that includes prehospital stroke screen results, to a Statewide Emergency Department (ED) Surveillance System database containing ED disposition diagnosis ICD-9 codes. We compared EMS stroke screen results to the ED ICD-9 codes for ischemic and hemorrhagic stroke and TIA. Results: Over a 24 month period, we identified 3,901 linked records with a prehospital impression of acute stroke. This included 2,419 (62%) records with a completed and conclusive stroke screen result. There were 1,202 patients (50%) with a completed CPSS and 1,217 patients (50%) with a completed LAPSS. Ninety-four EMS agencies from 65 of the state’s 100 counties were represented in the complete data. The CPSS was 80% (95% CI 76-83%) sensitive and 46% (95% CI 42-50%) specific, whereas the LAPSS was 73% (95% CI 69-76%) sensitive and 42% (95% CI 38-46%) specific for identifying patients with an ED diagnosis of stroke or TIA. Conclusion: When performed and conclusively recorded by many different EMS agencies across one state, the CPSS and LAPSS had similar test characteristics. If prehospital screening is to be used to determine transport diversion to acute stroke centers, improving the specificity of these screens would be optimum.

The feasibility and value of low-frequency data collected using colocated 2-Hz and 10-Hz geophones

ConocoPhillips acquired a production 3D surface seismic survey in 2010. The survey size was about 410 square miles recorded by 10-Hz geophones with INOVA 364 vibrators that are capable of sweeping from 1 to 150 Hz. In addition, we carried out a field experiment recording a swath of 3D surface seismic data with colocated 2-Hz and 10-Hz geophones. The cost of acquiring the 2-Hz geophone data was negligible when compared with the cost of the entire 3D survey. Because the vibrators can produce energy down to 2 Hz, the use of the 2-Hz geophone is crucial in capturing this energy. To our knowledge, this was the first field experiment employing the 2-Hz geophones and vibrators that could transmit enough low-frequency signals into the ground. The questions we investigate in this field experiment are: (1) how much low-frequency signal can be recorded using 2-Hz geophones, (2) how much degradation of low-frequency signal results from the 10-Hz geophones when compared to the 2-Hz geophones, and (3) the possibility of using the colocated data sets to enhance the low-frequency signal of 10-Hz geophone data that include both experimental and production data. The analyses of the 10-Hz and 2-Hz geophone data in prestack and poststack domains concluded that 2-Hz geophone data clearly exhibited more low-frequency signal than the 10-Hz geophone data. The 2-Hz geophone stack had low-frequency signal down to 2 Hz, and a spiking deconvolution further extended the amplitude spectrum down to 1 Hz. In addition, we develop a novel technique to derive a deterministic match filter from the colocated data sets. The application of this filter on the 10-Hz geophone data recovers the low-frequency signal below 10 Hz.

Trends in testing and notification for genital gonorrhoea in a northern Australian district, 2004 - 2008

The study aimed to examine the trends in notification and testing for genital gonorrhoea (Neisseria gonorrhoeae) in the Darwin Remote District of Northern Territory, Australia, between 2004 and 2008. Using laboratory testing data and notification data, we calculated the annual sex- and age-specific notification rates, testing rates and positivity rates, and examined their trends. A deterministic matching method was used to identify unique individuals tested in order to estimate the number of years out of five in which each individual was tested. The correlation between testing rates and notification rates was calculated. The notification rates for the 15-24 year age group increased sharply from 2004 to 2005, and then trended downwards between 2005 and 2008, with a decrease of 48.2% in females and 59.9% in males. No evident trends were found in testing rates. The positivity rates for this age group decreased by 46.3% in females (from 8.9% to 4.8%), and by 70.4% in males (from 10.8% to 3.2%) between 2004 and 2008. Over 76% of the population in this age-group had been tested at least once during the study period. A moderate correlation was found between notification rates and testing rates in both sexes. There was a significant decreasing trend in the notification rate of gonorrhoea between 2005 and 2008, which was most probably due to a decrease in prevalence. This study demonstrates the importance and utility of population-level testing data in understanding the epidemiology of common bacterial sexually transmissible infections such as gonorrhoea.

Adverse Birth Outcomes among Nulliparous vs. Multiparous Women

Previous studies indicate that nulliparous women (i.e., women having no previous births) are at higher risk for adverse birth outcomes than multiparous women (i.e., women having had at least one previous birth). We examined whether part of the difference in adverse outcome rates is attributable to nulliparous women with poor pregnancy outcomes being less likely (through choice or fecundity differences) to have a subsequent live birth within the same time period as nulliparous women without adverse outcomes. Using deterministic matching, we linked nulliparous women from the North Carolina Detailed Birth Record to subsequent births. We employed statistical and simulation-based analyses to estimate first birth outcome rate differences between nulliparous women who did have a subsequent live birth vs. those who did not. Our Markov simulations focused on preterm birth (PTB). Among nulliparous women who were not linked to a second birth, maternal age-adjusted rates of multiple adverse outcomes were all statistically higher compared with rates for linked women. These results also held in race/ethnicity-specific analyses. Simulations found that the relative risk of PTB associated with a history of PTB was underestimated if some women who would have been at risk for PTB did not experience a second birth. The observed differences in rates of adverse outcomes between nulliparous and multiparous women are partly attributable to higher-risk women not having a subsequent live birth, either by choice or due to fecundity differences.

A spectral matching quality indicator for material mapping using spectral library search methods

Spectral library search methods are being used increasingly as an efficient approach for exploiting hyperspectral remotely sensed data in material identification and mapping applications. The aim of this study was to develop a quantitative method, using an indicator called the Quality factor (Q-factor), for providing quantitative information on the reliability of spectral identifications in the interpretation (classification) of unknown spectra by library search methods. This was achieved by summing the two main requirements of a typical reflectance spectral library search for material mapping: (1) a reliable correlation between spectral matching scores and material similarity, and (2) a reliable separation ability between the relevant and non-relevant parts of the candidate reference spectra. These form a metric whose values reflect the closeness of the output reference spectra to the input unknown spectra for a chosen library search method. The Q-factor was tested as an indicator of the reliability of the material identifications by the library search for a range of unknown reflectance spectra of various types of vegetation, soils and minerals collected from the US Geological Survey (USGS) Spectral Library and from our in-house spectral database. The results indicate that this approach has the potential to separate correct and incorrect spectral identifications resulting from a particular spectral library search method using a reference similarity logic. The method may be applied to any combination of deterministic spectral matching alternatives using reflectance spectra. Spectrum-level quality information provided by the Q-factor is useful for optimizing a particular search method or for choosing the most appropriate method for distinct identification and classification problems.

A FUZZY-UPDATED CACHE OF AUTOMATA MATCHING FOR EMBEDDED NETWORK PROCESSOR

Light-weight network gateways often employ a cost-effective embedded network processor and have received a strong demand for empowering content filtering services. In this regard, we were motivated to propose a specialized cache, fuzzy-updated cache automata matching (FCAM) circuit for accelerating the embedded network processors. Although automata matching algorithms are robust with deterministic matching time, there is still plenty of room for improving its average-case performance. The proposed FCAM employs cache to accelerate the root state and nonroot state with the multiple characters matching, and applies the fuzzy decision to improve the cache performance. In our experiment, the FPGA implementation of FCAM can perform at the rate of 10.5 Giga bits per second with the patterns of 25,642 bytes. This performance is superior to previous matching hardware in terms of throughput and pattern set.

Use and cost of health services among overweight and obese Canadian children

Along with a dramatic rise in the rates of childhood obesity, obesity-related disorders, such as type 2 diabetes, hypertension, and obstructive sleep apnea, are seen with increasing frequency in children. As a consequence, overweight and obese children may use health care services more often than their normal weight peers. The aim of the current study was to assess health service use and costs across categories of weight status. Prospective cohort study using data from a population-based survey among grade 5 children in the Canadian province of Nova Scotia linked with administrative health data, using a combination of deterministic and probabalistic matching (n = 4 380). Total health care costs (physician and hospital costs), lifetime (up to age 14 years) physician costs and number of physician visits were assessed in a series of multiple regression models. There was a gradient for higher costs and utilization across the three weight groups. Total health care costs in the three years following the survey were 21% (95% CI: 2-43) higher in obese children compared with normal weight children. Obese children also had significantly higher lifetime physician costs and more physician visits than their normal weight peers. The health care cost trajectories of normal weight and obese children drift apart as early as 3 years of age. Interpretation. Obese children in the Canadian province of Nova Scotia have significantly higher health care costs and more physician visits and specialist referrals than their normal weight peers, highlighting the need for cost-effectiveness studies of obesity prevention programs.

Empirical aspects of record linkage across multiple data sets using statistical linkage keys: the experience of the PIAC cohort study

BackgroundIn Australia, many community service program data collections developed over the last decade, including several for aged care programs, contain a statistical linkage key (SLK) to enable derivation of client-level data. In addition, a common SLK is now used in many collections to facilitate the statistical examination of cross-program use. In 2005, the Pathways in Aged Care (PIAC) cohort study was funded to create a linked aged care database using the common SLK to enable analysis of pathways through aged care services.Linkage using an SLK is commonly deterministic. The purpose of this paper is to describe an extended deterministic record linkage strategy for situations where there is a general person identifier (e.g. an SLK) and several additional variables suitable for data linkage. This approach can allow for variation in client information recorded on different databases.MethodsA stepwise deterministic record linkage algorithm was developed to link datasets using an SLK and several other variables. Three measures of likely match accuracy were used: the discriminating power of match key values, an estimated false match rate, and an estimated step-specific trade-off between true and false matches. The method was validated through examining link properties and clerical review of three samples of links.ResultsThe deterministic algorithm resulted in up to an 11% increase in links compared with simple deterministic matching using an SLK. The links identified are of high quality: validation samples showed that less than 0.5% of links were false positives, and very few matches were made using non-unique match information (0.01%). There was a high degree of consistency in the characteristics of linked events.ConclusionsThe linkage strategy described in this paper has allowed the linking of multiple large aged care service datasets using a statistical linkage key while allowing for variation in its reporting. More widely, our deterministic algorithm, based on statistical properties of match keys, is a useful addition to the linker's toolkit. In particular, it may prove attractive when insufficient data are available for clerical review or follow-up, and the researcher has fewer options in relation to probabilistic linkage.

Using a linear array to estimate the velocity of underwater moving targets

A new method uses a linear array that takes advantage of underwater physical sound fields to estimate the velocity of an underwater moving target. The mathematical model was established by considering the geometric relationship between the moving target installed with only two transducers to radiate sound of different frequencies and the linear array. In addition, deterministic maximum likelihood and signal phase matching algorithms were introduced to effectively find the directions of arrival (DOAs) of the sound sources of the two transducers installed on the target. Factors causing velocity measurement errors were considered. To track the target, a linear array with a compass, a pressure transducer, a signal conditioner and a digital recorder was configured. Relevant requirements for the array parameters were derived. The simulation showed that a 16-element array with an aperture of less than 1m can measure velocity with relative error of no more than 4% when including typical system errors. Anechoic pool and reservoir experiments confirmed these results.

A fast scalable automaton-matching accelerator for embedded content processors

Home and office network gateways often employ a cost-effective embedded network processor to handle their network services. Such network gateways have received strong demand for applications dealing with intrusion detection, keyword blocking, antivirus and antispam. Accordingly, we were motivated to propose an appropriate fast scalable automaton-matching (FSAM) hardware to accelerate the embedded network processors. Although automaton matching algorithms are robust with deterministic matching time, there is still plenty of room for improving their average-case performance. FSAM employs novel prehash and root-index techniques to accelerate the matching for the nonroot states and the root state, respectively, in automation based hardware. The prehash approach uses some hashing functions to pretest the input substring for the nonroot states while the root-index approach handles multiple bytes in one single matching for the root state. Also, FSAM is applied in a prevalent automaton algorithm, Aho-Corasick (AC), which is often used in many content-filtering applications. When implemented in FPGA, FSAM can perform at the rate of 11.1Gbps with the pattern set of 32,634 bytes, demonstrating that our proposed approach can use a small logic circuit to achieve a competitive performance, although a larger memory is used. Furthermore, the amount of patterns in FSAM is not limited by the amount of internal circuits and memories. If the high-speed external memories are employed, FSAM can support up to 21,302 patterns while maintaining similar high performance.

The experience of linking Victorian emergency medical service trauma data

BackgroundThe linking of a large Emergency Medical Service (EMS) dataset with the Victorian Department of Human Services (DHS) hospital datasets and Victorian State Trauma Outcome Registry and Monitoring (VSTORM) dataset to determine patient outcomes has not previously been undertaken in Victoria. The objective of this study was to identify the linkage rate of a large EMS trauma dataset with the Department of Human Services hospital datasets and VSTORM dataset.MethodsThe linking of an EMS trauma dataset to the hospital datasets utilised deterministic and probabilistic matching. The linking of three EMS trauma datasets to the VSTORM dataset utilised deterministic, probabilistic and manual matching.ResultsThere were 66.7% of patients from the EMS dataset located in the VEMD. There were 96% of patients located in the VAED who were defined in the VEMD as being admitted to hospital. 3.7% of patients located in the VAED could not be found in the VEMD due to hospitals not reporting to the VEMD. For the EMS datasets, there was a 146% increase in successful links with the trauma profile dataset, a 221% increase in successful links with the mechanism of injury only dataset, and a 46% increase with sudden deterioration dataset, to VSTORM when using manual compared to deterministic matching.ConclusionThis study has demonstrated that EMS data can be successfully linked to other health related datasets using deterministic and probabilistic matching with varying levels of success. The quality of EMS data needs to be improved to ensure better linkage success rates with other health related datasets.

Linking Millennium Cohort data to birth registration and hospital episode records

The Millennium Cohort Study of 18,818 UK babies born in 2000-02 interviewed parents when the baby was 9 months old. Time constraints on the interview limited the amount of health-related questions that could be included. The aim of this study was to augment interview data with information from birth registrations and hospital records. It also provided an opportunity to assess the accuracy of the data acquired and parents' recall of the information on pregnancy and delivery. Deterministic and probabilistic matching were used to obtain information from birth registration and hospital records. Investigation into the accuracy of the matches obtained was undertaken. The records received were checked for range, consistency and completion. Birth registration data were obtained for 99% of those who gave consent. The number of additional variables gained ranged from six in Northern Ireland to 16 in Scotland. Hospital record data were obtained for 83% of those who gave consent. The additional general and maternity-related variables gained ranged from 55 in Scotland to 76 in England. Completion of available health record variables ranged from 28% to 100% across all UK countries. Linkage to birth registration and hospital records in order to augment Millennium Cohort Study data with routinely collected data was successful. The variables gained by linkage have added considerable value to the cohort study and validated some of the mother's responses.

Testing High-Resolution ADCs With Low-Resolution/Accuracy Deterministic Dynamic Element Matched DACs

This paper presents a deterministic dynamic element matching (DDEM) approach, which is applied to low-precision digital-to-analog converters (DACs) to generate uniformly spaced voltage samples for analog-to-digital converter (ADC) testing. Theoretical analysis is provided to show the test performance using this DDEM DAC. Both simulation results and experimental results from a fabricated DDEM DAC are presented to verify the performance. The ADC testing performance, by using an 8-bit DDEM DAC (linearity less than 5 bits without DDEM), is comparable to the best results reported in the literature using on-chip linear ramp generators. The DDEM technique offers great potential for use in both production test and built-in-self-test (BIST) environments.

Deterministic high-speed root-hashing automaton matching coprocessor for embedded network processor

While string matching plays an important role in deep packet inspection applications, its software algorithms are insufficient to meet the demands of high-speed performance. Accordingly, we were motivated to propose fast and deterministic performance root-hashing automaton matching (RHAM) coprocessor for embedded network processor. Although automaton algorithms are robust with deterministic matching time, there is still plenty of room for improvement of their average-case performance. The proposed RHAM employs novel root-hashing technique to accelerate automaton matching. In our experiment, RHAM is implemented in a prevalent automaton algorithm, Aho-Corasick (AC) which is often used in many packet inspection applications. Compared to the original AC, RHAM only requires extra vector size in 48 Kbytes for root-hashing, and has about 900% and 420% outperformance for 20,000 URLs and 10,000 virus patterns respectively. Implementaion of RHAM FPGA can perform at the rate of 12.6 Gbps with the pattern amount in 34,215 bytes. This is superior to all previous matching hardware in terms of throughput and pattern set.

Complete Pattern Matching for DNA Computing

Pattern matching is essential in many applications such as information retrieval, logic programming, theorem-proving, term rewriting and DNA-computing. It usually breaks down into two categories: root and complete pattern matching. Root matching determines whether a subject term is an instance of a pattern in a pattern set while complete matching determines whether a subject term contains a sub-term that is an instance of a pattern in a pattern set. For the sake of efficiency, root pattern matching need to be deterministic and lazy. Furthermore, complete pattern matching also needs to be parallel. Unlike root pattern matching, complete matching received little interest from the researchers of the field. In this paper, we present a novel deterministic multi-threaded complete matching method. This method subsumes a deterministic lazy root matching technique that was developped by the authors in an earlier work. We evaluate the performance of proposed method using theorem-proving and DNA-computing applications.

Protein−Ligand NOE Matching: A High-Throughput Method for Binding Pose Evaluation That Does Not Require Protein NMR Resonance Assignments

Given the three-dimensional (3D) structure of a protein, the binding pose of a ligand can be determined using distance restraints derived from assigned intra-ligand and protein-ligand nuclear Overhauser effects (NOEs). A primary limitation of this approach is the need for resonance assignments of the ligand-bound protein. We have developed an approach that utilizes data from 3D 13C-edited, 13C/15N-filtered HSQC-NOESY spectra for evaluating ligand binding poses without requiring protein NMR resonance assignments. Only the 1H NMR assignments of the bound ligand are essential. Trial ligand binding poses are generated by any suitable method (e.g., computational docking). For each trial binding pose, the 3D 13C-edited, 13C/15N-filtered HSQC-NOESY spectrum is predicted, and the predicted and observed patterns of protein-ligand NOEs are matched and scored using a fast, deterministic bipartite graph matching algorithm. The best scoring (lowest "cost") poses are identified. Our method can incorporate any explicit restraints or protein assignment data that are available, and many extensions of the basic procedure are feasible. Only a single sample is required, and the method can be applied to both slowly and rapidly exchanging ligands. The method was applied to three test cases: one complex involving muscle fatty acid-binding protein (mFABP) and two complexes involving the leukocyte function-associated antigen 1 (LFA-1) I-domain. Without using experimental protein NMR assignments, the method identified the known binding poses with good accuracy. The addition of experimental protein NMR assignments improves the results. Our "NOE matching" approach is expected to be widely applicable; i.e., it does not appear to depend on a fortuitous distribution of binding pocket residues.

Validation of Probabilistic Linkage to Match De‐identified Ambulance Records to a State Trauma Registry

To validate the accuracy of using probabilistic linkage for matching de-identified ambulance records to a state trauma registry. This was a retrospective cohort analysis. Three thousand nine hundred nineteen true matches between ambulance and state trauma registry data from 1998 to 2003 were identified by deterministic matching on trauma identification number and verified by human review. Two thousand thirty-eight ambulance records from trauma patients not meeting criteria for a true match, and an identical number of trauma registry records randomly selected from the one local county served by a different EMS provider, were included as nonmatches. There were 17 variables considered for linkage, which included the following: age, gender, race, county, hospital, date, rural setting, call and arrival times, mechanism, penetrating injury, vital signs, intubation, and intoxication. Probabilistic linkage was used to link the two data sets, using seven different combinations of common variables (maximum, 17; minimum, 4). The sensitivity and specificity of identifying true matches and nonmatches (95% confidence intervals [95% CI]) were calculated for each combination of variables. Using all 17 available variables, 3,766 of 3,919 true matches were appropriately linked (sensitivity, 96.1%; 95% CI = 95.4% to 96.7%), with eight mismatches (specificity, 99.6%; 95% CI = 99.2% to 99.8%). Sensitivity fell below 95% with < 15 variables; however, sensitivity was very dependent on the inclusion of variables with high discriminatory power. Specificity remained >98% regardless of the number of variables included. Probabilistic linkage is a valid method for matching ambulance records to a trauma registry without the use of patient identifiers; however, the sensitivity of identifying true matches is critically dependent on the number and type of common variables included in the analysis.