Chinese
Thank you for visiting Nature.com.The browser version you are using has limited support for CSS.For the best experience, we recommend that you use an updated browser (or turn off compatibility mode in Internet Explorer).In the meantime, to ensure continued support, we will display the site without styles and JavaScript.
Fluid intake is important to prevent dehydration and reduce recurrent kidney stones.There has been a trend in recent years to develop tools for monitoring fluid intake using “smart” products such as smart bottles.There are several commercially smart baby bottles available, mainly aimed at health-conscious adults.To our knowledge, these bottles have not been validated in the literature.This study compared the performance and functionality of four commercially available smart feeding bottles.The bottles are H2OPal, HidrateSpark Steel, HidrateSpark 3 and Thermos Smart Lid.One hundred ingestion events per bottle were recorded and analyzed and compared to ground truth obtained from high-resolution scales.H2OPal has the lowest mean percent error (MPE) and is able to balance errors across multiple sips.HidrateSpark 3 provides the most consistent and reliable results with the lowest sip errors per time.The MPE values of the HidrateSpark bottles were further improved using linear regression as they had more consistent individual error values.The Thermos Smart Lid was the least accurate, as the sensor did not extend across the entire bottle, causing many records to be lost.
Dehydration is a very serious problem because it can lead to adverse complications, including confusion, falls, hospitalization, and death.Fluid intake balance is important, especially in older adults and people with underlying medical conditions that affect fluid regulation.Patients at risk of recurrent stone formation are advised to consume large amounts of fluids.Therefore, monitoring fluid intake is a useful method to determine if adequate fluid intake is being taken1,2.There are many attempts in the literature to create reports of systems or devices that can help track and manage fluid intake.Unfortunately, most of these studies did not result in a commercially available product.Bottles on the market are primarily aimed at recreational athletes or health-conscious adults looking to add hydration.In this article, we aimed to determine whether common, commercially available water bottles are a viable solution for researchers and patients.We compared four commercial water bottles in terms of performance and functionality.The bottles are HidrateSpark 34, HidrateSpark Steel5, H2O Pal6 and Thermos Smart Lid7 as shown in Figure 1.These bottles were chosen because they are one of the only four popular bottles that are (1) available for purchase in Canada and (2) have sip volume data accessible through the mobile app.
Images of analyzed commercial bottles: (a) HidrateSpark 34, (b) HidrateSpark Steel5, (c) H2OPal6, (d) Thermos Smart Lid7.The red dashed box shows the location of the sensor.
Of the above bottles, only previous versions of HidrateSpark have been validated in research8.The study found that the HidrateSpark bottle was accurate within 3% of measuring total intake over a 24-hour period of fluid intake.HidrateSpark has also been used in clinical studies to monitor intake in patients with kidney stones9.Since then, HidrateSpark has developed new bottles with different sensors.H2OPal has been used in other studies to track and promote fluid intake, but no specific studies have validated its performance2,10.Pletcher et al.The geriatric features and information available online were compared for several commercial bottles, but they did not perform any validation of their accuracy11.
All four commercial bottles include a free proprietary app for displaying and storing ingestion events transmitted via Bluetooth.The HidrateSpark 3 and Thermos Smart Lid have the sensor in the middle of the bottle, possibly using a capacitive sensor, while the HidrateSpark Steel and H2Opal have a sensor on the bottom, using a load or pressure sensor.The sensor location is shown in the red dashed box in Figure 1.In the Thermos Smart Lid, the sensor cannot reach the bottom of the container.
Each bottle is tested in two phases: (1) a controlled suction phase and (2) a free-living phase.In both phases, the results recorded by the bottle (obtained from the product mobile app used on Android 11) were compared with ground truth obtained using a 5 kg scale (Starfrit Electronic Kitchen Scale 93756).All bottles were calibrated before data was collected using the app.In Phase 1, sip sizes from 10 mL to 100 mL of 10 mL to 100 mL were measured in random order, 5 measurements each, for a total of 50 measurements per vial.These events are not actual drinking events in humans, but are poured out so that the amount of each sip can be better controlled.At this stage, recalibrate the bottle if the sip error is greater than 50 mL, and re-pair if the app loses the bluetooth connection to the bottle.During the free-life phase, a user drinks water freely from a bottle during the day, and they choose different sips.This phase also includes 50 sips over time, but not all of them in a row.Therefore, each bottle has a dataset of a total of 100 measurements.
To determine total fluid intake and ensure proper daily hydration, it is more important to have accurate volumetric intake measurements throughout the day (24 hours) rather than every sip.However, to identify prompt intervention cues, each sip needs to have a low error, as was done in the study by Conroy et al. 2 .If the sip is not recorded or recorded poorly, it is crucial that the bottle can balance the volume on the next recording.Therefore, the error (measured volume – actual volume) is adjusted manually.For example, suppose the subject drank 10 mL and the bottle reported 0 mL, but then the subject drank 20 mL and the bottle reported a total of 30 mL, the adjusted error would be 0 mL.
Table 1 lists various performance metrics for each bottle considering two phases (100 sips).The mean percent error (MPE) per sip, mean absolute error (MAE) per sip, and cumulative MPE are calculated as follows:
where \({S}_{act}^{i}\) and \({S}_{est}^{i}\) are the actual and estimated intakes of \({i}_{th}\ ) sip, and \(n\) is the total number of sips.\({C}_{act}^{k}\) and \({C}_{est}^{k}\) represent the cumulative intake of the last \(k\) sips.The Sip MPE looks at the percent error for each individual sip, while the Cumulative MPE looks at the total percent error over time.According to the results in Table 1, H2OPal has the lowest number of lost records, the lowest Sip MPE, and the lowest cumulative MPE.The mean error is better than the mean absolute error (MAE) as a comparison metric when determining total intake over time.Because it illustrates the bottle’s ability to recover from poor measurements over time while recording subsequent measurements.The sip MAE is also included in applications where the accuracy of each sip is important because it calculates the absolute error of each sip.Cumulative MPE also measures how well the measurements are balanced across the phase and does not penalize a single sip.Another observation was that 3 of the 4 bottles underestimated the volume intake per mouth shown in Table 1 with negative numbers.
The R-squared Pearson correlation coefficients for all bottles are also shown in Table 1.HidrateSpark 3 provides the highest correlation coefficient.Although HidrateSpark 3 has some missing records, most of them are small mouths (The Bland-Altman plot in Figure 2 also confirms that HidrateSpark 3 has the smallest limit of agreement (LoA) compared to the other three bottles.LoA analyzes how well the actual and measured values agree.Furthermore, almost all measurements were in the LoA range, which confirms that this bottle provides consistent results, as shown in Figure 2c.However, most values are below zero, which means that the size of sip is often underestimated.The same is true for HidrateSpark Steel in Figure 2b, where most of the error values are negative.Therefore, these two bottles provide the highest MPE and cumulative MPE compared to H2Opal and Thermos Smart Lid, with errors distributed above and below 0, as shown in Fig. 2a,d.
Bland-Altman plots of (a) H2OPal, (b) HidrateSpark Steel, (c) HidrateSpark 3 and (d) Thermos Smart Lid.The dashed line represents the confidence interval around the mean, calculated from the standard deviation in Table 1.
HidrateSpark Steel and H2OPal had similar standard deviations of 20.04 mL and 21.41 mL, respectively.Figures 2a,b also show that the values of HidrateSpark Steel always bounce around the mean, but generally stay within the LoA region, while H2Opal has more values outside the LoA region.The maximum standard deviation of the Thermos Smart Lid was 35.42 mL, and more than 10% of the measurements were outside the LoA region shown in Figure 2d.This bottle provided the smallest Sip Mean Error and relatively small Cumulative MPE, despite having the most missing records and the largest standard deviation.The Thermos SmartLid has a lot of missed recordings because the sensor straw does not extend to the bottom of the container, causing missed recordings when the liquid content is below the sensor stick (~80 mL).This should lead to an underestimation of fluid intake; however, Thermos was the only bottle with positive MPE and Sip Mean Error, implying that the bottle overestimated fluid intake.So, the reason Thermos’ average sip error is so low is because the measurement is overestimated for nearly every bottle.When these overestimates are averaged, including many missed sips that are not recorded at all (or “underestimated”), the average result is balanced.When excluding missed records from the calculation, the Sip Mean Error became +10.38 mL, confirming a large overestimation of a single sip.While this may seem positive, the bottle is actually inaccurate in individual sip estimates and unreliable because it misses many drinking events.Furthermore, as shown in Figure 2d, Thermos SmartLid seems to increase the error with increasing sip size.
All in all, H2OPal was the most accurate at estimating sips over time, and the most reliable way to measure most recordings.The Thermos Smart Lid was the least accurate and missed more sips than the other bottles.The HidrateSpark 3 bottle had more consistent error values, but underestimated most sips that resulted in poor performance over time.
It turns out that the bottle may have some offset that can be compensated for using a calibration algorithm.This is especially true for the HidrateSpark bottle, which has a small standard deviation of error and always underestimates a single sip.A least squares (LS) method was used with stage 1 data while excluding any missing records to obtain offset and gain values.The resulting equation was used for the sip intake measured in the second stage to calculate the actual value and to determine the calibrated error.Table 2 shows that calibration improved the Sip mean error for two HidrateSpark bottles, but not H2OPal or Thermos Smart Lid.
During Phase 1 where all measurements are done, each bottle is refilled multiple times, so the calculated MAE may be affected by the bottle fill level.To determine this, each bottle is divided into three levels, high, medium, and low, based on the total volume of each bottle.For the Phase 1 measurements, a one-way ANOVA test was performed to determine whether the levels were significantly different in absolute error.For HidrateSpark 3 and Steel, the errors for the three categories are not significantly different.There was a borderline significant difference (p Two-tailed t-tests were performed to compare stage 1 and stage 2 errors for each bottle.We achieved p > 0.05 for all bottles, which means that the two groups were not significantly different.However, it was observed that the two HidrateSpark bottles lost a much higher number of recordings in stage 2.For H2OPal, the number of missed recordings was nearly equal (2 vs. 3), while for Thermos SmartLid there were fewer missed recordings (6 vs. 10).Since the HidrateSpark bottles were all improved after calibration, a t-test was also performed after calibration.For HidrateSpark 3, there is a significant difference in errors between Stage 1 and Stage 2 (p = 0.046).This is more likely due to the higher number of missing records in stage 2 compared to stage 1.
This section provides insights into the usability of the bottle and its application, as well as other functional information.While bottle accuracy is important, the usability factor is also important when choosing a bottle.
HidrateSpark 3 and HidrateSpark Steel are equipped with LED lights that remind users to drink water if they don’t meet their goals as planned, or flash a certain number of times per day (set by the user).They can also be set to flash every time the user drinks.H2OPal and Thermos Smart Lid do not have any visual feedback to remind users to drink water.However, all purchased bottles have mobile notifications to remind users to drink through the mobile app.The number of notifications per day can be customized in the HidrateSpark and H2OPal applications.
HidrateSpark 3 and Steel use linear trends to guide users when to drink water and give an hourly suggested goal that users should hit by the end of the day.H2OPal and Thermos Smart Lid only provide a daily total goal.In all bottles, if the device is not connected to the app via bluetooth, the data will be stored locally and synced after pairing.
None of the four bottles focuses on hydration for seniors.Additionally, the formulas the bottles use to determine daily intake goals are not available, making it difficult to determine whether they are suitable for older adults.Most of these bottles are large and heavy and not tailored for seniors.The use of mobile apps may also not be ideal for older adults, although it may be useful for researchers to collect data remotely.
All bottles cannot determine if the liquid has been consumed, discarded or spilled.All bottles also need to be placed on a surface after each sip to accurately record intake.This means that drinks may be missed if the bottle is not set down, especially when refilling.
Another limitation is that the device needs to be periodically re-paired with the app to sync data.The Thermos needed to be re-paired every time the app was opened, and the HidrateSpark bottle often struggled to find a Bluetooth connection.H2OPal is easiest to re-pair with the app if the connection is lost.All bottles are calibrated before testing begins and must be recalibrated at least once during the process.The HidrateSpark bottle and H2OPal must be emptied and completely filled for calibration.
All bottles do not have the option to download or save data long-term.Also, none of them can be accessed through the API.
HidrateSpark 3 and H2OPal use replaceable lithium-ion batteries, HidrateSpark Steel and Thermos SmartLid use rechargeable batteries.As stated by the manufacturer, the rechargeable battery should last up to 2 weeks on a full charge, however, it must be recharged almost weekly when using the Thermos SmartLid heavily.This is a limitation as many people won’t remember to recharge the bottle regularly.
There are a variety of factors that can influence the choice of a smart bottle, especially when the user is an elderly person.The weight and volume of the bottle is an important factor as it needs to be easy to use by frail seniors.As mentioned earlier, these bottles are not tailored for seniors.The price and quantity of liquid per bottle is also another factor.Table 3 shows the height, weight, liquid volume and price of each bottle.Thermos Smart Lid is the cheapest and lightest as it is made entirely of lighter plastic.It also holds the most liquids compared to the other three bottles.Conversely, H2OPal was the tallest, heaviest and most expensive of the research bottles.
Commercially available smart bottles are useful for researchers because there is no need to prototype new devices.Although there are many smart water bottles available, the most common problem is that users do not have access to the data or raw signals, and only some results are displayed in the mobile app.There is a need to develop a widely used smart bottle with high accuracy and fully accessible data, especially one tailored for the elderly.Out of the four bottles tested, H2OPal out of the box had the lowest Sip MPE, cumulative MPE, and number of missed recordings.HidrateSpark 3 has the highest linearity, smallest standard deviation and lowest MAE.HidrateSpark Steel and HidrateSpark 3 can simply be manually calibrated to reduce the Sip mean error using the LS method.For more accurate sip recordings, the HidrateSpark 3 is the bottle of choice, while for more consistent measurements over time, the H2OPal is the first choice.The Thermos SmartLid had the least reliable performance, had the most missed sips, and overestimated individual sips.
The study is not without limitations.In real-world scenarios, many users will drink from other containers, especially hot liquids, store-bought beverages, and alcohol.Future work should evaluate how each bottle’s form factor affects errors to guide smart water bottle design.
Rule, AD, Lieske, JC & Pais, VM Jr. 2020. Kidney Stone Management.JAMA 323, 1961–1962.https://doi.org/10.1001/jama.2020.0662 (2020).
Conroy, DE, West, AB, Brunke-Reese, D., Thomaz, E. & Streeper, NM Timely adaptive intervention to promote fluid consumption in patients with kidney stones.Health Psychology.39, 1062 (2020).
Cohen, R., Fernie, G., and Roshan Fekr, A. Fluid intake monitoring systems in the elderly: a literature review.Nutrients 13, 2092. https://doi.org/10.3390/nu13062092 (2021).
Inc, H. HidrateSpark 3 Smart Water Bottle & Free Hydration Tracker App – Black https://hidratespark.com/products/black-hidrate-spark-3. Accessed April 21, 2021.
HidrateSpark STEEL Insulated Stainless Steel Smart Water Bottle and App – Hidrate Inc. https://hidratespark.com/products/hidratespark-steel.Accessed April 21, 2021.
Thermos® Connected Hydration Bottle with Smart Cap.https://www.thermos.com/smartlid.Accessed on November 9, 2020.
Borofsky, MS, Dauw, CA, York, N., Terry, C. & Lingeman, JE Accuracy of measuring daily fluid intake using a “smart” water bottle.Urolithiasis 46, 343–348.https://doi.org/10.1007/s00240-017-1006-x (2018).
Bernard, J., Song, L., Henderson, B. & Tasian, GE. Association between daily water intake and 24-hour urine output in adolescents with kidney stones.Urology 140, 150–154.https://doi.org/10.1016/j.urology.2020.01.024 (2020).
Fallmann, S., Psychoula, I., Chen, L., Chen, F., Doyle, J., Triboan, D. Reality and perception: Activity monitoring and data collection in real-world smart homes.In the 2017 IEEE SmartWorld Conference Proceedings, Ubiquitous Intelligence and Computing, Advanced and Trusted Computing, Scalable Computing and Communications, Cloud and Big Data Computing, Internet of People and Smart City Innovation (SmartWorld/SCALCOM/UIC/ATC/ CBDCom/IOP/SCI ), 1-6 (IEEE, 2017).
Pletcher, DA et al.An interactive water drinking gadget designed for the elderly and Alzheimer’s patients.In a lawsuit on the human side of IT for the elderly population.Social Media, Games, and Assisted Environments (eds Zhou, J. & Salvendy, G.) 444–463 (Springer International Publishing, 2019).
This work was supported by a Canadian Institutes of Health Research (CIHR) Foundation Grant (FDN-148450).Dr. Fernie received the funding as the Creaghan Chair of Family Prevention and Medical Technology.
Kite Institute, Toronto Rehabilitation Institute – University Health Network, Toronto, Canada
Conceptualization – RC; Methodology – RC, AR; Writing – Manuscript Preparation – RC, AR; Writing – Review and Editing, GF, AR; Supervision – AR, GF All authors have read and agree with the manuscript published version.
Springer Nature remains neutral with regard to jurisdictional claims of published maps and institutional affiliations.
Open Access This article is licensed under the Creative Commons Attribution 4.0 International License, which allows use, sharing, adaptation, distribution, and reproduction in any medium or format, provided you give proper credit to the original author and source, providing a Creative Commons license , and indicate whether changes have been made.Images or other third-party materials in this article are included under the Creative Commons license of the article, unless otherwise noted in the credits for the material.If the material is not included in the Creative Commons license of the article and your intended use is not permitted by law or regulation or exceeds that permitted, you will need to obtain permission directly from the copyright owner.To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
Cohen, R., Fernie, G., and Roshan Fekr, A. Monitoring liquid intake in commercially available smart water bottles.Science Rep 12, 4402 (2022).https://doi.org/10.1038/s41598-022-08335-5
By submitting a comment, you agree to abide by our Terms and Community Guidelines.If you see abusive content or content that doesn’t comply with our terms or guidelines, please flag it as inappropriate.
Post time: Mar-29-2022
